Organizational Routines: Advancing Empirical Research

Organizational Routines Organizational Routines Advancing Empirical Research Edited by Markus C. Becker Professor, S...

Author: Marcus C. Becker

34 downloads 1235 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Organizational Routines

Organizational Routines Advancing Empirical Research

Edited by

Markus C. Becker Professor, Strategic Organization Design Unit, Department of Marketing and Management, University of Southern Denmark

Nathalie Lazaric Researcher, GREDEG, CNRS, University of Nice-Sophia Antipolis, France

Edward Elgar Cheltenham, UK • Northampton, MA, USA

© Markus C. Becker and Nathalie Lazaric 2009 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA

A catalogue record for this book is available from the British Library Library of Congress Control Number: 2008943826

ISBN 978 1 84720 194 2 Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall

Contents List of contributors 1

Advancing empirical research on organizational routines: introduction Markus C. Becker and Nathalie Lazaric

PART I 2 3

5

6

7

8

11 26

METHODS FOR ANALYSING ORGANIZATIONAL ROUTINES

Using workflow data to explore the structure of an organizational routine Brian T. Pentland, Thorvald Haerem and Derek W. Hillison The contribution of event-sequence analysis to the study of organizational routines Carlo Salvato The inheritance of organizational routines and the emergence of a firm genealogy in the fashion design industry Rik Wenting

PART III

1

CONCEPTUAL FOUNDATIONS

Routines as technologies and as organizational capabilities Richard R. Nelson The nature and replication of routines Geoﬀrey M. Hodgson

PART II

4

vii

47

68

103

ORGANIZATIONAL ROUTINES AND STABILITY IN ORGANIZATIONS

Is it the ‘same’? Observing the regeneration of organizational character at Camp Poplar Grove Jeremy P. Birnholtz, Michael D. Cohen and Susannah V. Hoch Uncovering inertia: ambiguity between formal rules and routines of interaction Martijn van der Steen v

131

159

Contents

vi

PART IV

9

10

11

ORGANIZATIONAL ROUTINES AND ORGANIZATIONAL CHANGE AND INNOVATION

The influence of artefacts and distributed agencies on routines’ dynamics: from representation to performation Luciana D’Adderio Innovation routines: exploring the role of procedures and stable behaviour patterns in innovation Markus C. Becker and Francesco Zirpoli The diﬃcult creation of novel routines: persistence of old habits and renewal of knowledge base in French SMEs Frédéric Huet and Nathalie Lazaric

Index

185

223

248

277

Contributors Markus C. Becker holds a PhD in Management from the Judge Business School, Cambridge University. He has held positions with the Centre National de Recherche Scientifique (CNRS) at the University of Strasbourg, France, and with the University of Southern Denmark. He is currently Professor of Organization Theory at the Strategic Organization Design Unit, Department of Marketing and Management, University of Southern Denmark. Jeremy P. Birnholtz is Assistant Professor in the Department of Communication and the Faculty of Computing and Information Science at Cornell University. Jeremy received his PhD from the School of Information at the University of Michigan in 2005, and is interested in improving the usefulness and usability of collaboration technologies through a focus on human attention, and in the intersections of social/ behavioural science theory and technology design. Michael D. Cohen is the William D. Hamilton Professor of Complex Systems, Information and Public Policy at the University of Michigan. He has written a number of studies on routine using both experimental methods and field observations. He is a co-author of ‘A garbage can model of organizational choice’ (with James March and Johan Olsen, 1972) and of Harnessing Complexity (with Robert Axelrod, 1999). Luciana D’Adderio is a Lecturer in Technology Management at the University of Edinburgh, UK and Innovation Fellow with the Advanced Institute of Management Research (AIM). As part of her Fellowship she is currently developing a new research area on ‘Dependable innovation in global manufacturing’. She is the author of the book Inside the Virtual Product: How Organizations Create Knowledge through Software published by Edward Elgar. Thorvald Haerem holds a PhD in Management from Copenhagen Business School. He is Associate Professor of Organizational Psychology in the Department of Leadership and Organizational Management at the Norwegian School of Management. He has published his work in journals such as Journal of Applied Psychology, International Journal of Organizational Analysis and Scandinavian Library Research. vii

viii

Contributors

Derek W. Hillison is a PhD candidate in the Department of Accounting and Information Systems at Michigan State University. He holds an MBA and a BS in Management Information Systems from Florida State University. His research interests include workflow, the management of business process, organizational routines, and the development of information systems. Geoﬀrey M. Hodgson is Research Professor in Business Studies at the University of Hertfordshire in England. He has published several books – including Economics in the Shadows of Darwin and Marx (2006), The Evolution of Institutional Economics (2004) and How Economics Forgot History (2001) – and over 100 articles in academic journals. He is a Fellow of the Academy of Social Sciences and Editor-in-Chief of the Journal of Institutional Economics. His website is www.geoﬀrey-hodgson.info. Susannah V. Hoch is a Software Engineer for Kwajalein Range Services at the Ronald Reagan Ballistic Missile Defense Test Site. She designs, develops and evaluates Graphical User Interfaces for various decisionmaking systems. She received her MS in Human–Computer Interaction from the School of Information at the University of Michigan and her BA in Psychology from the University of Pennsylvania. Frédéric Huet is Assistant Professor at University of Compiègne, France. His research fields include the economics of organization and innovation. More specifically, his work focuses on cooperative relationships as the new predominant model of organization in the knowledge-based economy. Investigations have recently considered the impact/causation of the digital paradigm and convergence on such relationships as well as their territorial embeddedness. Nathalie Lazaric is Researcher at Centre National de Recherche Scientifique (CNRS), University of Nice-Sophia Antipolis, France. Her research interests include evolutionary theory and organizational routines; articulation and codification of knowledge; and knowledge produced in diverse organizational set-ups. She has co-edited several books, including Trust and Economic Learning (1998) and Knowledge, Learning and Routines (2003). Her publications have appeared in Industrial and Corporate Change; Research Policy; Journal of Evolutionary Economics and elsewhere. Richard R. Nelson is the Henry R. Luce Professor of International Political Economy at Columbia University. His research has concentrated on the processes of long-run economic change, with particular emphasis on technological advances and on the evolution of economic institutions. Some of his publications include The Sources of Economic Growth (Harvard, 2000),

Contributors

ix

Sources of Industrial Leadership (Cambridge, 1999), National Innovation Systems: A Comparative Analysis (Oxford, 1993), An Evolutionary Theory of Economic Change (Harvard 1985, reprint), and many others. Brian T. Pentland is Professor in the Department of Accounting and Information Systems at Michigan State University. He received his PhD in Management from the Massachusetts Institute of Technology in 1991. His publications have appeared in Academy of Management Review; Accounting, Organizations and Society; Administrative Science Quarterly; Industrial and Corporate Change; Management Science; Organization Science and elsewhere. Carlo Salvato is Associate Professor of Strategic Management at Bocconi University, Milan, where he received a PhD in Business Administration and Management. He also received a PhD in Strategy and Entrepreneurship from Jönköping International Business School, Sweden, where he is currently Research Fellow at the Center for Family Enterprise and Ownership. His research interests and publications focus on the evolution of routines and capabilities, and on the entrepreneurial evolution of private and closely held firms. Martijn van der Steen completed his PhD at the University of Groningen, where he currently works as an Assistant Professor in Accounting. His research interests include the ways in which management accounting can induce behavioural changes and the role of agency in changes in management accounting systems. In addition, Martijn investigates how management accounting influences organizational sensemaking. Rik Wenting has studied economics and geography at Utrecht University and business administration at Erasmus University, Rotterdam. He graduated from Utrecht University in 2004 with a Masters thesis on the spatial evolution of the British automobile industry. Subsequently, Wenting started a PhD project at Utrecht University on the spatial evolution of creative industries. He has published articles on the economic geography of industrial dynamics in several international journals. Francesco Zirpoli gained his MPhil and PhD in Management Studies from the University of Cambridge, UK, and his Doctorate in Business Administration from the University of Naples, Italy. He is Lecturer with tenure at the Department of Mechanical Engineering of the University of Salerno and is currently Fulbright Visiting Scholar at Columbia University. His research focuses on innovation management, technology and organization, strategic organization design, innovation networks and design outsourcing strategy.

1.

Advancing empirical research on organizational routines: introduction1 Markus C. Becker and Nathalie Lazaric

Since the concept of organizational routines was popularized by Nelson and Winter (1982), its potential to understand organizations and organizational change has attracted scholars’ attention. Organizational routines hold one of the keys to understanding how organizations accomplish their tasks, how they change, and how organizational capabilities are accumulated, transferred and applied (Cohen et al., 1996; Winter, 2000). Most of the potential of the concept of organizational routines to contribute to understanding organizations and how they change is still untapped, however. The last years, though, have seen increased empirical research that applies the concept of organizational routines in understanding organizations. Furthermore, diﬀerent methodologies for empirical research on organizational routines have also been pioneered, described and tested recently. Results of empirical studies employing these methodologies are starting to accumulate. The present volume collects a number of empirical contributions that apply the concept of organization routines for casting light on organizational phenomena. It showcases the state of the art in empirical research on organizations which employs the concept of organizational routines in understanding organizations. This volume follows up on the special section ‘Towards an operationalization of the routines concept’ in Industrial and Corporate Change, 14(5).2 Consulting that section and the present volume will show how the chapters in this volume tackle some of the challenges in applying organizational routines in analysing organizations that were described in the special section (Becker et al., 2005). We leave it to readers to judge how far empirical research applying the concept of organizational routines has already come in realizing the potential of the organizational routines concept to understanding organizations. The present volume is a display of the fact, however, that such research is moving forward, and that issues of empirical research on routines are currently being tackled. In the remainder of this introductory 1

2

Advancing empirical research on organizational routines: introduction

chapter, we sketch briefly which issues the chapters speak to, and how they are connected. We then conclude by a brief outlook on further empirical research on organizational routines. The two chapters in Part I provide diﬀerent perspectives on the conceptual foundations of the routines concept. Richard Nelson provides a perspective on organizational routines that emphasizes routines as ‘technologies’, that is, productive techniques for doing something. In a ‘physical technology’, physical inputs, apparatus and procedures employing these, largely control what is done, with the organizational aspects perhaps complex but largely derivative of the technology. In a ‘social technology’, on the other hand, human interactions are the dominant part of the story, with the physical aspects of the technology relatively simple. As Nelson explains in his chapter, the operation of organizational routines tends to involve both physical and social technologies. Organizational capabilities involve both its ‘engineering’ strength, in the sense of command of the physical technologies it needs to operate, and its strength in management and ability to coordinate the various activities it needs to employ to be eﬀective. Taking such a perspective allows asking fruitful research questions, such as how and why a routine works as it does, how routines evolve, and what the role of interaction is between social and physical technologies. Chapters such as the ones by Luciana D’Adderio and Brian Pentland, Thorvald Haerem and Derek Hillison are consistent with such a view and provide illustration of it. In his chapter, Geoﬀrey Hodgson proposes a perspective on organizational routines that sees them as dispositions to energize conditional patterns of behaviour within an organized group of individuals, involving sequential responses to cues. Such a perspective helps distinguish organizational routines from individual habits. Routines are structures of interlocking individual habits, and thus more than mere aggregations of individual-level habits. The chapter also discusses how routines act as repositories of knowledge and how they replicate. One of the key promises of the organizational routines concept is to provide a means of describing organizational behaviour and how it changes. To have an empirical science of organizational routines, however, we need to be able to identify particular routines and compare them to other routines. There are two issues that need to be tackled in order to advance here. Few quantitative methods are available for unpacking the structure and evolution of routines. Moreover, which data should be used for such an exercise? As to the first issue, one of the most promising methods so far is event-sequence analysis, pioneered by Brian Pentland. The two chapters in Part II represent an introduction to the method. They go beyond a simple introduction, however. They also provide two specific examples of how it can be applied, and what insights are to be garnered by applying

Advancing empirical research on organizational routines: introduction

3

it. Carlo Salvato’s chapter oﬀers a step-by-step guide of how to apply the method, while Pentland, Haerem and Hillison oﬀer a particularly careful and comprehensive overview of techniques for analysing and representing the structure of organizational routines, and a discussion of the methodological and practical issues involved in such a method. As to the problem of data, the two chapters show how workflow event logs (Pentland, Haerem and Hillison) and archival material on organizational processes (Salvato) can be used to analyse organizational routines. The chapters illustrate these techniques using over 2000 performances of an invoice approval routine, as captured in a workflow management system (Pentland, Haerem and Hillison), and archival data on development processes of 90 products initiated by Alessi, the design firm (Salvato). These two chapters provide a great opportunity for advancing the possibility of drawing on empirical research on routines to cast light on issues such as the impact of rules and procedures on behaviour patterns (see D’Adderio, Becker and Zirpoli, and van der Steen in this volume), how to select the appropriate level of granularity for describing routines, and how to advance our understanding of routine combinatorics, one of the basic ways in which organizations evolve (Becker et al., 2006). Rik Wenting’s chapter focuses on describing a diﬀerent aspect, that is, the replication of routines. He studies the transfer of organizational routines from parent firms to their spin-oﬀ companies in the fashion industry from its inception in 1858 to today. Drawing on a unique dataset of the fashion design industry covering 554 biographies of the world’s top fashion designers, Wenting traces inheritance back to all previous employers of the entrepreneur, constituting a complete genealogical graph. He then tests whether quantity or quality of pre-entry experiences of the founder aﬀected firm survival. His article makes a contribution on methods to study the replication of routines. It also casts light on the eﬀects of routine replication, which he studies over an unusually long time horizon. We just mentioned how empirical research on organizational routines can advance our understanding of routine replication and routine combinatorics, and thus, organizational evolution. The chapters also inform two other fundamental questions that organizational routines can cast light on; the sources of stability and change in organizations. The chapters in Part III contribute to understanding stability in organizations by analysing organizational routines. Jeremy Birnholtz, Michael Cohen and Susannah Hoch present a qualitative case study that analyses sources of stability in organizations that lie dormant for substantial periods and are regenerated after such dormant periods. What are the sources of stability that make such organizations recognizably ‘the same’? Summer camps provide a unique environment for exploring these issues because they lie dormant for much of the year until their members come together to bring

4

Advancing empirical research on organizational routines: introduction

the organization ‘back to life’, where the result of this regeneration process is recognizably another instance of ‘the same’ organization to repeat clientele. The chapter also informs the question regarding how we can refer to a routine as ‘the same’ entity, even when there are substantial diﬀerences in diﬀerent specific instances. It also represents one of the few empirical articles that shed light on how to identify dispositions, and how such mutually adapted action dispositions have an impact by forming a coherent system. The chapter by Martijn van der Steen focuses on rules as sources of stability in behaviour patterns. Analysing the case of a change initiative at Rabobank, he investigates the interaction between the formal rules as embedded in the accounting and controlling system, and behaviour patterns. His chapter contributes to understanding under which conditions such interaction contributes to inertia, and under which circumstances it contributes to organizational change. The chapter also casts light on the impact of incentives and management tools on the interaction between rules and action patterns (cf. the chapters by D’Adderio and Becker and Zirpoli in this volume). Part IV groups chapters that cast light on organizational routines as drivers of organizational change and novelty. In her chapter, Luciana D’Adderio builds on recent advances in Routines and Performativity Theory to provide a nuanced characterization of the key influence of artefacts and agencies on the change and stability of routines. Ultimately, in this framework, a routine’s stable pattern emerges out of a double recursive loop of convergence and divergence between procedure and performances and the competitive organizational arrangements in which the routine is stabilized. Drawing from the ethnographical observation of the ‘freeze’ routine at a leading automotive manufacturer, this chapter thus shows how some organizational agencies are able to inscribe their abstract views of the routine in artefacts (that is SOPs, rules) and tools (that is software), thus creating stronger agencements. These heterogeneous agentic programmes, being deeply embedded in a thick web of organizational relationships, are more diﬃcult to oppose and thus can ‘turn exceptions into rules’. The balance between stabilization and destabilization is not, therefore, the result of the interplay between structures, on one side, and agencies, on the other, but the outcome of the struggles among competing performative programmes. The stability resulting from performative struggles is similar to Nelson and Winter’s (1982) notion of ‘truce’ as a continuously challenged and emergent achievement; this framework, however, captures the direction, intensity and persistence of the forces that are responsible for stabilizing or destabilizing the routine. The chapter by Markus Becker and Francesco Zirpoli investigates the role of stable behaviour patterns and of standard operating procedures that can contribute to the stability of behaviour patterns in innovation processes. They start from the observation that

Advancing empirical research on organizational routines: introduction

5

procedures and stable behaviour patterns for accomplishing innovation tasks have very prominent roles in the new product development management literature, and that many innovating firms have many such rules and procedures and carry out innovation in routinized ways. This also includes examples of radical innovation, such as at IDEO, the design firm. Following this lead, they ask ‘what is the role, if any, of standard operating procedures and stable behaviour patterns in repeatedly generating innovations?’ The question is pursued with a case study of the use of virtual simulation technology at the Engineering Center of a major European car manufacturer. Frédéric Huet and Nathalie Lazaric focus on the organizational conservatism of firms and the reluctance to regenerate their knowledge base. In their chapter, they explain why firms are reluctant to change their organizational routines. By discussing the notion of transactions and cooperation, they identify diverse ‘routine transactions’ which are related to habitual activities involving stabilized knowledge, and ‘strategic transactions’ related to situations of novelty implying new opportunities and for which there is no stabilized knowledge. This terminology, developed by Commons (1950), is applied in explaining organizational conservatism in a population of SMEs which are pushed to innovate to survive in hostile environments but are reluctant to do so, because implementing the creation of novel routines implies a revision of their representations (that is their beliefs or their taken-for-granted assumptions). So far, we have briefly described what questions the chapters tackle, and what contributions they make to these questions. In so doing, the chapters also push the frontier by identifying new questions that can guide further research.

QUESTIONS ON METHODS What level of granularity to select for describing routines is a critical methodological issue. How do we select a level of granularity that assures we describe organizational routines that are meaningful (Becker et al., 2005)? Recent empirical research has marshalled econometric measures for coping with the task of describing organizational variation present in routines, such as variety in how tasks are accomplished (Pentland, Haerem and Hillison), in the content of interactions between firms (Huet and Lazaric), or of antecedents of the actual routines (Wenting). All these measures present progress regarding our ability to describe and identify organizational variation. Because such methods are now available, and describing and identifying organizational variation has become possible, how to decide on the appropriate level of granularity for such a description has become even

6

Advancing empirical research on organizational routines: introduction

more important than before. As analytical methods have become available, we might want to scrutinize them for whether they have an in-built bias for particular levels of granularity (either in the method itself or in the data that is available to be analysed with a particular method).

QUESTIONS ON STABILITY AND CHANGE The chapters have taken three diﬀerent perspectives for understanding the stability or change of behaviour patterns for achieving organizational tasks: routine combinatorics, replication, and rules and standard operating procedures as sources of stability or change. In each, they have not just added to our understanding but have also flagged new questions that can guide further research. Regarding rules and standard operating procedures, some authors in this volume and in the literature more generally are inspired by the dialectics and recursive interaction between the description of behaviour (ostensive) and the behaviour itself (performative) (Feldman and Pentland, 2003). It appears that much current empirical research is exploring this line of inquiry. One important impulse provided by Feldman and Pentland (2003) was to emphasize that rules and operating procedures do not always have just a stabilizing eﬀect on behaviour patterns. Rather, under certain conditions they can also lead to endogenous organizational change or to novel outcomes of organizational processes. Identifying such conditions would appear a natural next step in this line of inquiry. As for routine combinatorics, describing the building blocks that are recombined seems to be the next issue that needs to be solved. Deciding on what are meaningful levels of granularity is obviously important here. Once the building blocks are identified, identifying the mechanisms of recombination is high on the research agenda. As Hodgson argued in his chapter, stable behaviour patterns are emergent properties of generative structures. Regarding replication, an important item on the research agenda is therefore to describe the replication of these generative structures that generate stable behaviour patterns, not of the behaviour patterns themselves. The generative structures are diﬃcult to observe, yet hold the key to understanding the replication of stable behaviour patterns. Wenting’s chapter does a nice job in focusing on such generative structures as the object of replication, rather than specific behaviour patterns. Another issue put on the research agenda by Wenting’s chapter and other emergent research is the role of space in replication. While economic geography has long insisted on proximity as an important variable explaining the results of replication processes, integrating space in models of the replication of organizational routines is still an unaccomplished task.

Advancing empirical research on organizational routines: introduction

7

In conclusion, the chapters in the present volume oﬀer insights from empirical research that show how the organizational routines concept can advance our understanding of organizations and how they change. Much still remains to be done. The volume provides evidence, however, that empirical research employing the organizational routines concept in analysing organizations is under way and has increased over the last few years. Importantly, such empirical research also starts to be more cumulative as it begins to follow the same lines of inquiry and to use the same methods. These are all hallmarks of an advancing field.

NOTES 1. The work on this volume has benefited from a research grant from the Agence Nationale de Recherche (ANR) under the ‘Jeunes Chercheuses et Jeunes Chercheurs’ programme (grant no. JC05_44029). 2. It also provides a companion to the Handbook of Organizational Routines (Becker, 2008).

REFERENCES Becker, Markus C. (ed.) (2008), Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Becker, Markus C., Thorbjørn Knudsen and James G. March (2006), ‘Schumpeter, Winter, and the sources of novelty’, Industrial and Corporate Change, 15(2), 353–71. Becker, Markus C., Nathalie Lazaric, Richard R. Nelson and Sidney G. Winter (2005), ‘Applying organizational routines in analyzing organizations’, Industrial and Corporate Change, 14(5), 775–91. Cohen, Michael D., Roger Burkhart, Giovanni Dosi, Massimo Egidi, Luigi Marengo, Massimo Warglien and Sidney G. Winter (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5(3), 653–98. Commons, J.R. (1950), The Economics of Collective Action, Madison, WI: University of Wisconsin Press. Feldman, M.S. and B.T. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48, 94–118. Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Belknap Press of Harvard University Press. Winter, Sidney G. (2000), ‘The satisficing principle in capability learning’, Strategic Management Journal, 21, 981–96.

PART I

Conceptual foundations

2.

Routines as technologies and as organizational capabilities Richard R. Nelson

INTRODUCTION The focal theme of the conference at which the chapters in this volume were first presented was on the relationship between the behavioral and cognitive aspects of routines. I interpret the cognitive aspects as being about understanding how and why a routine works as it does, as contrasted with the know-how that is required simply to ‘do it’, that is the behavioral knowledge base. Also, I would argue that the former – the understanding – tends to be articulated to a considerable extent, while the latter – the ability to do – can be largely tacit. The theme of this chapter is that the importance, and the power, of the understanding bearing on routines varies greatly across diﬀerent kinds of routines. The diﬀerences matter profoundly in terms of how diﬀerent routines, or aspects of routines, evolve. Let me begin my discussion by reminding you that in our book Winter and I (1982) highlighted two diﬀerent aspects of a ‘routine’. From one point of view, a routine is a ‘technology’. By that we did not mean that it necessarily involved fancy equipment or materials, but rather that it was a productive technique for doing something, as a program, or a recipe (including the steps that transform the inputs into the intended output), that could be described without specifying any particular way the actions required by the technique were to be assigned to particular individuals and groups, and coordinated. Many ‘technologies’, for example the way teachers try to teach children how to read, do not involve fancy Technology with a capital T. Highly relevant to the discussion at this conference, neither do many ways of doing things that tend to be called ‘business practice’. Other technologies are very complex and sophisticated, are described in engineering journals and texts, and require significant study and learning-by-doing by young engineers and scientists before they are mastered. From another point of view a routine is an organizational capability. By this we did not mean to underplay either the often complex and sophisticated technological aspect of some routines, that to a considerable extent 11

12

Conceptual foundations

involves shared knowledge of a technological community that transcends the individual organizations that employ that technology, or the role of the skills and knowledge required of individual humans who were involved in the ‘doing’. Rather we wanted to highlight that most productive routines required the coordinated actions of a number of individuals. Developing an eﬀective team operation of a routine often is a challenging problem, and once achieved, is one often requiring maintenance. In that sense, mastery of a complex routine is an organizational capability. These two facets were combined in our concept of a routine. On the one hand, a routine is defined in terms of a collection of actions in appropriate sequence that gets the job done, sometimes with closely specified inputs, and machinery designed to process those inputs in a particular way, in other cases with little of the above. On the other hand, a routine is an activity involving a number of people and groups, coordinated by particular patterns of cooperative interaction, with often many aspects under explicit management direction. There are major advantages in seeing these two aspects as strongly complementary, rather than separate. In many cases it is virtually impossible to consider them separately. Thus the Ford mass production ‘routine’ for making automobiles, introduced during the 1920s, involved a tightly linked combination of machinery, job specifications, and mechanisms for coordination and management. Similar was the Toyota system of lean production introduced during the 1980s. On the other hand, some routines are dominated by the ‘technological’ aspect, as in continuous flow production of chemicals, or processes where particular artifacts largely structure what is done. The examples I will develop later are corn production using hybrid seeds, and polio vaccination. Others largely involve human interaction and coordination, with the technological aspect being relatively simple. The examples I will develop are procedures for improving production operations through quality circles, and managerial decision making under the M form of organization. Bhaven Sampat and I have coined the terms ‘physical’ and ‘social’ technologies to connote routines where diﬀerent aspects dominate (Nelson and Sampat, 2001). In a ‘physical technology’ physical inputs, apparatus, and procedures employing these largely control what is done, with the organizational aspects perhaps complex but largely derivative of the technology. On the other hand in a ‘social technology’ human interactions are the dominant part of the story, with the physical aspects of the technology being relatively simple. In many cases routines can not easily be classified as one or the other, with both the physical and the social aspects being complex and important. However, in many cases it does seem reasonable to classify a routine as largely the one or the other.

Routines as technologies and as organizational capabilities

13

In any case, it is clear that the theory about routines and their evolution that Winter and I developed has in fact spawned two quite separate intellectual communities. One is concerned with organizational capabilities and behavior, where the focus is on ‘social’ technologies. The other is concerned with the evolution of ‘physical’ technologies. It also is evident that these two diﬀerent communities look at the role of understanding in the evolution of routines in diﬀerent ways. The community focused on organizations sees routines as largely activities involving coordinated actions of individuals and groups, with little of Technology with a capital T involved. The evolution of routines is seen as largely driven by experience, with some but generally limited understanding on the part of the actors regarding how to go about deliberately improving routines systematically. The community focused on technological advance sees the activities involved in doing something – producing a product or service – as involving particular inputs and equipment, with improvement of technologies a deliberate continuing process, and with scientific and technological understanding providing powerful guidance to the process. The basic diﬀerences here, Sampat and I have proposed, are between the characteristics of physical and social technologies.

HOW DO ROUTINES EVOLVE? But before getting into the diﬀerences, it is important to get clear on some general aspects of how routines evolve. Winter and I made the argument that extant routines must be understood as having evolved partly as a counter to the theory that what individuals and organizations do should be understood as the consequence of optimizing choices. Our counter proposal was that the ways of doing things that one observes in operation at any time generally are in place as a result of a stream of decisions and actions, each one somewhat myopic. And regarding the details of that dynamic process, we proposed that it was illuminating to try to understand change over time in the range of routines employed as being driven by a process involving variation and selection. However, beyond the common aspect of both having a dynamic involving variation, and selection, we did not intend to connote that the evolution of routines was ‘just like’ the evolution of biological species. Unfortunately in my view, there has been a tendency of social scientists who adopt the broad evolutionary idea to presume that a theory of the evolution of routines must look a lot like evolutionary theory in biology. I think it important, therefore, to lay out some of the key diﬀerences. One very important diﬀerence between the evolution of routines and

14

Conceptual foundations

biological evolution is that, unlike phenotypes who have no choice about the genes they have, firms can and do make decisions regarding the routines they have in place. There is no unanimity within the community of scholars studying how significant changes in the routines used by organizations in a broad economic sector come about. Almost always such change involves the death of some older organizations and the birth of new ones. But in the transitions I have studied, part of the change, in some cases the lion’s share, involves firms switching their routines. That is, changes in the importance of diﬀerent routines involves changes in what particular firms do, as well as changes in the composition of extant firms. This is very diﬀerent from in biological evolution. A related diﬀerence is the important role played by human purpose, deliberation and decision making, in the processes through which productive routines evolve. These factors play key roles both in the generation of new alternatives for consideration, and in the selection process. This is not to argue that the routines in place at any time are there because specific decisions were made that they should be in place. As I noted earlier, the rationality in Nelson–Winter evolutionary theory is very much a bounded rationality, and what is in place at any time is almost never the result of some earlier plan that has been eﬀected in detail. However, human and organizational decision making is very much part of the evolutionary processes that result in the routines that are in use. One important consequence is that the relevant ‘variation’ needs to be looked at in a diﬀerent way. Variation in Darwinian biological evolution is variation in genes, and traits and behaviors, in an extant population at any time. This is the ‘stuﬀ’ on which selection works. However, in the evolution of productive ways of doing things used by human organizations a good portion of the relevant variation is in human minds, and is explored through analysis, discussion and argument, rather than in actual practice. Thus a team of engineers contemplate a wide range of plausible designs, and gradually home in on one, before they actually build and release a new product for use. Managers of a business firm may contemplate a range of possible actions before choosing one and putting it to practice. The actual variation at any time tends to be a small part of the contemplated variation, and an important part of the selection process involves the winnowing down of alternative ideas for action before actual action is taken. Again, I do not mean here that calculations, or judgments, regarding what is the appropriate thing to do are reliable guides to action. In some circumstances they may be no better than flipping coins. However, in other circumstances understanding may be quite strong, and analysis and judgment may be very eﬀective in pointing action in the right direction, and in helping disastrous actions to be avoided. And in any case, it is clear that

Routines as technologies and as organizational capabilities

15

the attempts of human individuals and organizations to think through what they ought to be doing are an important part of the evolutionary process. Third, while the organizations in a sector may be competitive in many respects, there are important elements of shared understanding and information. The sharing may be informal and spontaneous. But in many cases there are formal structures that facilitate and shape the sharing, industry and professional associations for example. There are trade journals. Business School professors and faculty at Engineering or Medical Schools kibitz in the processes through which evaluations of new routines are made, and often play a central role in those processes. This collective aspect of the evolution of routines is another factor that diﬀerentiates the process here from evolution in biology. Given these characteristics of the way routines evolve, what are the conditions under which one would expect to see rapid and cumulatively great advances in the performance of the practices employed for particular purposes, and conversely, when these conditions are absent, one would expect slow and sporadic progress? First of all, routines, or aspects of routines, diﬀer in the extent to which they in fact are ‘routinized’, and I want to propose that a certain degree of routinization is itself an important condition for a practice to be improved systematically. The actors involved need to be able, at the least, to hold a practice in place if that is the intention, and to replicate it with some degree of fidelity if they so choose. Since, as I have suggested above, the advance of practice is to some extent a collective process, rapid sustained improvement requires that one actor be able to take advantage of what others have learned, and this requires that the practice be imitable to some degree. Second, there must be a reasonably strong and consistent eﬀect of using or not using a routine that matters to the individuals or organizations that can use it. This is required to the extent that the evolution of techniques in use depends on who survives, who expands, and who declines or dies. Since, as argued above, whether a user sticks with the routine it has been using, or switches to another one, generally is a matter of conscious decision, beliefs about the eﬃcacy of a routine matter. These beliefs at any time may closely reflect the actual eﬀectiveness of the diﬀerent available routines, or they may not. I am not a hard core ‘social constructionist’. Therefore I believe that beliefs that are not consistent with the reality tend to be fragile. Thus what I am arguing here is that rapid improvement of a routine or the development and adoption of better ones depends to a considerable degree on the ability of users to correctly evaluate the eﬀectiveness of diﬀerent routines.

16

Conceptual foundations

I want to highlight that condition one is necessary for condition two. If there is considerable variation over time in what the routine actually is, one cannot expect a consistent result of using ‘it’. If there is considerable variation in the details of the routine across users, the experience of one will provide little useful information regarding how the routine (broadly defined) will work for another user. However, the holding of the first condition obviously is not suﬃcient for the second to hold. Third, the actors need to be able to learn from natural or intentional experiments, and to be able to bring to practice what they have learned. I note that the first two conditions above are necessary for this third condition. To be able to learn and take advantage from variation, the actor needs to be able to hold and perhaps replicate the variants that are improvements over the status quo, and reject those that are worse. I note that these conditions permit a continuing process of learning-bydoing and using, perhaps supplemented by some deliberate experimentation to explore changes that for some reason seem promising. If there are a number of diﬀerent actors, or if some actors operate multiple activities, what is learned in one instance can be carried over to extensive use. Fourth, the opportunity cost of deliberate experimentation can be reduced significantly if the experimentation can be done eﬀectively oﬀ-line on a low cost ‘model’ of the actual routine. The model may be a theoretical model, or a physical model. In any case, for this route to advancement to be possible, what is learned experimenting with the model must provide reliable information about what will work with the real routine. All four of these conditions are tied to the strength of understanding about a practice and how and why it works as it does. In the first place, when such understanding is strong, the key aspects of a process that need to be kept in place in order for it to work properly are evident. Absent such understanding, it is not clear what holding a practice constant really entails. Second, and this is particularly important when the eﬃcacy of a practice can be known directly only after a considerable amount of time and experience, understanding can be of the form of reliable proxies that can be observed relatively quickly and cheaply. Third, understanding of why a particular practice works and how it works points to experiments that have promise of identifying likely improvements. In the absence of some such understanding, experimentation is literally ‘blind’ and productive only by luck. Fourth, when it is strong, understanding can lead to the development of parsimonious but reliable means of oﬀ-line learning. Absent a degree of understanding, modeling, either through theoretical calculation or simplified artifact, is not possible. For all of these reasons, the ability to continuously improve a body of practice depends on the strength of the underlying understanding.

Routines as technologies and as organizational capabilities

17

FOUR CASES In the appendix to this chapter I briefly consider four cases, two of the evolution of physical technologies – hybrid corn and polio vaccines – and two of the evolution of social technologies – quality circles, and the M form of organization. The diﬀerences between the two sets are striking. First of all, regarding the origins of the practices in question. In the case both of hybrid corn and polio vaccines, the new technologies were created through an oﬀ-line R&D process. In the case of the two social technologies, while there almost certainly were memos and discussions before the new practices were tried, there was nothing like an oﬀ-line R&D exploration involved. Rather, the key actors simply moved to try to put in place a new way of doing things. There were similarly dramatic diﬀerences in the processes of evaluation of the new ‘technologies’. Both the new hybrid corn seeds and polio vaccines were tested in a controlled setting before they were put into mass practice. Once put into practice, individual users, and the user community as a whole, were able to generate detailed information regarding eﬃcacy, in which they could have confidence. In the case of the two social technologies, deliberate sophisticated eﬀorts at evaluation were non-existent or rare. Studies by academics were and are not particularly convincing. The eﬃcacy of these practices remains a matter of individual judgment. Both of the new physical technologies clearly were major advances over prior practice. Further, they usefully can be considered to be part of a steady sequence of advances in their broad fields. There have been a sequence of advances in seeds that, together, have led to cumulatively dramatic advances in farm productivity. Similarly, the development of polio vaccines should be seen as part and parcel of a broad endeavor that has led over the years to the eradication of many diseases that used to plague humankind. In contrast, the evolution of modes of business organization and decision making has been subject to fads, as well as clear-cut advances, and progress in this general area has not been particularly impressive.

THE PRESENCE OR ABSENCE OF STRONG UNDERSTANDING: THE KEY DIFFERENCE IN THE CASES I want to propose that the key diﬀerence between these cases was the strength of understanding in the cases of the two physical technologies, and the weakness of understanding in the cases of the social technologies. Particular hybrid seed varieties and particular vaccines were made so

18

Conceptual foundations

that they were essentially uniform across instances. Thus it was clear what a particular seed or vaccine actually was. In both cases the artifacts were designed to be uniform. To achieve this result took a lot of research, experimentation and control, which were guided by relatively sophisticated understanding of the process of hybridization on the one hand, and of how to develop a vaccine to cope with a particular virus on the other hand. A key to success in both instances was knowledge about what was essential and what was peripheral about the artifacts on question. No batch of seed, or of vaccine, can be literally identical. What mattered was uniformity in the relevant dimensions, and knowing these was the key to establishing practical uniformity. This knowledge, plus knowledge of the external conditions that influenced the performance of a seed or vaccine, permitted the bulk of the searching for an eﬀective artifact to proceed ‘oﬀ-line’, in an experimental setting. Once an artifact was found that worked well in an experimental setting, this knowledge led to appropriate confidence that what worked in an experimental setting would with high probability work in the field or population where its use was intended, if the relevant external conditions (for example the program for fertilizing and watering crops, or giving vaccine) were established and maintained. And the uniformity of the artifact and that knowledge meant that users as a group could learn from the experience of the early users. I note that most of this knowledge was scientific ‘know-how’ knowledge, as contrasted with deep theory. However, it was well articulated and taught to and shared by scientists in the fields. In contrast, in the cases of the two social technologies, there was little in the way of understanding, or research technique, that clearly identified what was relevant and what was peripheral, to the two routines in question. A strong case can be made that achieving uniformity of social technologies is inherently more diﬃcult than for physical technologies. But with little or no knowledge of what was essential to performance and what was not, there was (and is) little guidance regarding what aspects to consider closely, and try to control. In both of these case, it is clear that there was considerable variation in the routine both across organizations purporting to use it, and over time within particular organizations. This made evaluation extremely diﬃcult. There always was the question of just what was being evaluated. For similar reasons, this made it very diﬃcult for firms to learn from each other. And clearly there were limited opportunities to learn from controlled experiments, oﬀ-line or on-line. I propose that the diﬀerences here are exactly those that lie behind the very diﬀerent perspectives on the nature of ‘routines’, and how they evolve,

Routines as technologies and as organizational capabilities

19

taken in the writings on organizational capabilities on the one hand, and in the literature on technology and technological advance on the other hand. Thus a recent paper by Zollo and Winter (2002) on ‘Deliberate learning and the evolution of dynamic capabilities’ considers dynamic capabilities of firms as the joint result of relatively unsystematic learning from experience, and more specific eﬀorts to articulate, codify and reflect on experience. The empirical arena they use as an example is capabilities for the management of strategic alliances, presumably including decision making regarding whether a new alliance should be established, as well as how it is structured and managed. For a scholar of technological advance and the management of the process, the tale they tell is rather like the Sherlock Holmes case of the barking dog. There is no discussion whatsoever of the strength of knowledge of management scientists about matters like when making a strategic alliance is appropriate and how to make such alliances work; one can draw from the absence of such discussion the judgment of the authors (and the firms they studied?) that such knowledge is so weak it is not worth talking about. Nor is there any discussion of the kinds of studies that can fruitfully be undertaken to decide whether a particular proposed alliance, and its proposed form, is a good idea or not. Again, the implicit assumption is that such studies do not gain the firm very much, over what it already knows on the basis of reflections on its experience, codified as well as tacit. What a contrast with the picture drawn by Prencipe, Davies, Hobday and their colleagues, in their book on The Business of Systems Integration (2003). The orientation here is to company decision making regarding the multi-component product systems (like automobiles and aircraft) that they design and put together. Major resources are put into the studies that lead up to the initiation of the design and development of a new system. In the course of the development of the system, there is a considerable amount of testing to evaluate the progress being made, and to identify the problems that are emerging. I propose that the basic diﬀerence is that an alliance is a social technology, and a product system is a physical technology, and that the strength of the understanding bearing on the former, which sheds light on how to proceed with decision making, is much weaker than the understanding bearing on the latter. And of course the consequences of this are not simply the short-run diﬀerences in how decisions are made and learning is gained. There are major diﬀerences in how the nature of business alliances, and product systems like aircraft evolve. I would propose that while the relevant business community has learned something over the years, it is not clear that today they know much more about what kind of alliances pay oﬀ, and how

20

Conceptual foundations

to structure alliances, than they did a number of years ago. In contrast, there has been steady and cumulative great advances over the years in the performance of product systems like aircraft. These observations of course raise questions about why human knowledge bearing on physical technologies is so much stronger than human knowledge bearing on social technologies. That is a fascinating question, but not one for this chapter.

APPENDIX: PHYSICAL AND SOCIAL TECHNOLOGIES AND THEIR EVOLUTION I summarize here four relatively well studied cases of innovation, two of them relatively clearly ‘physical’ technologies: hybrid corn and polio vaccines, and two of them clearly ‘social’ technologies: quality circles and the M form. The discussion of these cases is presented in somewhat more detail in my paper with Bhaven Sampat and Alexander Peterhansl (2004), ‘Why and how innovations get adopted: a tale of four models’. Hybrid Corn My first example is the production of corn using hybrid seeds. Hybrid corn seed first became available commercially in the 1920s. This achievement was the result of sustained eﬀorts of many agricultural research workers to develop a better corn seed. The search was facilitated by an increasingly strong scientific understanding of plant biology. In the search, ‘better’ was quite well defined, and meant more profitable to grow, all things considered. These other things to be considered included the opportunity cost of the land and labor involved in growing hybrid corn, the fertilizers and insecticides involved, and other particular characteristics involved in the use of hybrid seeds, as contrasted with the alternatives. The seed itself was a clearly defined thing. And the various activities needed of the farmer involved either the seed itself, or things that needed to be done for or to the plants that grew from the seed. These activities could be routinized to a considerable degree, and could be described in terms of routines to be followed. Because of the high degree of specificity and routinization involved in using hybrid seeds, the results of growing these hybrids in plots in experimentation stations provided credible evidence to farmers as to the results they would be likely to achieve if they used hybrids, and took care of their crops in the way recommended by the experimentation stations. Similarly, the experience of one farmer in using hybrids provided highly

Routines as technologies and as organizational capabilities

21

relevant information to his neighbors, who could if they wished buy exactly the same brand of seed, and grow their crops using basically the same techniques. Hybrids originally were made available in only a small number of regions. The particular characteristics of the seed itself needed to be carefully tailored to the soil type and climatic characteristics of the region. A seed that worked well in one region would not necessarily work well in another. However, the success of hybrid seeds in the original regions made agricultural scientists, and seed growers, confident that they could develop hybrids that were suitable to other regions. The techniques of developing, and testing, new hybrids were by then well understood, and indeed quite routine. Hybrids quite rapidly were developed for a wide range of regions, and by the 1950s were used by virtually all farmers who grew corn. And the same techniques of creating new and better hybrids in public and private experimentation stations that led to the development of specialized hybrids tailored to the conditions of a wide variety of diﬀerent regions, also led to the steady improvements of hybrids within each of the regions. The result of all this was a tremendous increase between the 1920s and the 1960s in the total factor productivity of corn production in the United States. Polio Vaccines I turn now to a second example: the virtual elimination of polio due to the development of polio vaccines. This development depended upon identification of the virus involved in polio, or in this case three diﬀerent viruses, which occurred in 1949. The broad set of ‘physical’ techniques for developing a vaccine for infectious illness, once the bacteria or virus causing that illness had been identified, long had been generally known. The underlying sciences and the associated techniques were strong. Thus identification of the viruses set in train several diﬀerent attempts to develop an eﬀective vaccine for polio. In the early 1950s Jonas Salk developed what he thought was a promising vaccine using killed virus. Based on that development, an extensive double-blind set of clinical sets was conducted, and the results clearly showed that the incidence of polio for the treated groups was substantially lower than in the controls. Note that the generally accepted criteria here were very sharp: reduced incidence of the disease, and no noticeable negative side eﬀects. Notice also that the technology itself was well defined. It was a particular vaccine. In the late 1950s Albert Sabin developed a diﬀerent vaccine for polio, which used live but attenuated virus. Previous experience with vaccines

22

Conceptual foundations

for other diseases indicated that attenuated virus vaccines tended to give longer immunity than killed virus vaccines, but there was a risk that the former itself could cause infection. Tests of the Sabin vaccine showed that it was no more dangerous to use than the Salk vaccine. And in addition to the high probability giving longer immunity, the Sabin also had the advantage of being able to be taken orally. In most countries use of the Sabin vaccine quickly replaced use of the Salk vaccine. Again note that the criteria were clear and sharp. The fact that the Sabin vaccine could be taken orally was in eﬀect an extra benefit of a physical technology that clearly rated as superior to its predecessor on the generally accepted selected criteria. And note also that the alternatives being evaluated here were sharply diﬀerent. One used killed virus, and the other attenuated virus. The sharpness of the criteria, the specificity of the technological alternatives, and the strength of the underlying scientific knowledge, clearly greatly facilitated oﬄine research, and experiments on non-human animals. However, the ultimate comparison was made in a controlled experiment online. And the specificity of the technologies, and the clearness of the criteria, also made it possible for many professionals to be involved in the developments, in one way or another. Quality Circles Turning now to our first ‘social technology’ – the employment of quality circles as a vehicle for improving firm operations – the history is so incredibly diﬀerent. Quality circles became fashionable as a ‘social technology’ to be used within business firms during the early and mid-1980s, a period of time when Japanese firms were doing extremely well in competition with American firms in certain industries, and American firms and scholars in management were trying to understand the sources of the Japanese prowess, so that American firms could adapt and adopt. At a very high level of abstraction, the proposition that using quality circles was eﬃcacious could be posed in terms of the long-run profitability of the enterprise. However, the theory about why the use of quality circles would prove profitable for the enterprise involved an amalgam of propositions, some focused directly on the notion that they would help to improve product and process ‘quality’ and through that route improve profitability, and some posed more broadly in terms of the desirability of getting more worker involvement, both objective and subjective, in guiding the management of the enterprise. The core of the quality circle concept generally was agreed to be the regular meeting of small groups of workers who together would try to get clear on how current practices aﬀect product quality, and to identify changes that would enhance product quality. But it was widely

Routines as technologies and as organizational capabilities

23

recognized that how individual firms went about doing this did, and probably should, diﬀer significantly from firm to firm. During the early 1980s there was a surge of articles, some focused on the Japanese experience, and many oriented toward analysis of American firms which had adopted quality circles. These articles recognized that having a quality circle meant diﬀerent things in diﬀerent firms. Nonetheless, the orientation was toward trying to evaluate the quantity and quality of the improvements engendered through quality circles. A number of the studies were content with the identification and description of certain specific improvements that they thought could be ascribed to the work of quality circles. Others aimed to do a statistical analysis. Almost all of the latter studies yielded no statistically significant results. In a way this is not surprising because, on the one hand, the independent variable certainly was very heterogeneous, and on the other hand, there was no sharp agreement on exactly how the eﬀectiveness of quality circles should be measured. Partly at least because of the absence of real evidence that having a quality circle was eﬃcacious, various studies show a significant decline in the number of articles reporting on or advocating quality circles after the mid-1980s. And it is highly probable that the number of firms who had organized structures that they ‘called’ quality circles diminished as well. On the other hand, several scholars have proposed that quality circles were part of a general set of developments regarding managerial practice that has moved things in the right direction. Thus Sidney Winter (in Cole and Scott, 2000) has argued that the package of proclivities associated with greater concern on the part of firms with the quality of their products, combined with mechanisms to pay detailed attention to the processes going on in the firm that aﬀected product quality, and involvement in the analysis of workers who were part of those processes, undoubtedly was a good thing. Cole and Scott (2000) basically endorse this position by asking the following: ‘Is it possible that some of the early publicized failures identified by academic researchers, such as quality circles, actually laid the groundwork for some of the later successes? If so, this would suggest that organizational researchers might well examine management fads in a broader time and space framework’ So there very well may have been some real progress made here in an evolving ‘social technology’. But if so, the process certainly has been very diﬀerent from that associated with progress based on a physical technology, and while direct comparison is impossible, because no one still knows how to measure the eﬀect of better quality management, it is a very safe bet that the cumulative progress achieved under the regime of hybrid corn technology or of polio vaccines has been enormously greater than the cumulative advance achieved under quality management.

24

Conceptual foundations

The M Form As a second example of a new social technology, I want to consider the development and spread of the M form as a structure to enable better management of large complex firms. As Alfred Chandler (1977) tells the story, the M form came into existence in the early part of the twentieth century, as a number of firms who had increased the range of products they produced or the number of markets in which they were active struggled with the management problems of enlarged scale and scope. The standard form of business organization at that time involved a number of functional departments, such as finance, marketing and production, which in turn reported to top management. The enlargement of the variety of things that firms were doing meant that each of these departments now had to deal with a more heterogeneous, as well as a larger, set of things. The idea for the M form was to divide the company up into divisions, with each division having full responsibility for a small range of products and one or a few markets. Each of the divisions would then have their own functional departments, who now would be able to deal with matters of smaller scale and scope. The function of top management was to monitor the divisions. It would seem apparent that the adoption of the M form would reduce the complexity and variety of the tasks faced by middle-level management. It also would reorient, and to some extent make more coherent, the task of top management. But there was the question of ‘the bottom line’; to what extent would the adoption of the M form actually increase firm eﬃciency and profitability? It was and is virtually impossible to confidently assess the real eﬃciency advantages associated with the adoption of the M form for two reasons. First, unlike a polio vaccine, or the use of a hybrid corn seed, controlled experimentation was and is impossible in this case. And with so many other things going on in the environment, and often in the firm, there is no real way that the eﬀect on profitability of adoption of the M form in and of itself can reliably be estimated. Second, several studies have shown clearly that the M form was diﬀerent things in diﬀerent companies, and even within a given company tended to change over time. For these reasons, adherence to the notion that large diversified firms should manage through the M form became, and remains, something that is justified to a considerable extent as a matter of faith, and logical argument that is much more abstract and simplified than the actual contexts in which firms operate. It would seem that something like the M form remains the standard mode for organizing large and diversified business. Without doubt, individual businesses learned some things through their experiences with their

Routines as technologies and as organizational capabilities

25

organizational and managerial structures that have enabled them to eradicate some of the serious bugs. But to my knowledge, there exists no serious research which has argued that the M form, as a way of organizing large complex businesses, has been significantly refined and improved over the years so that the structures it has today are significantly more eﬃcient than the structures associated with the M form twenty years ago, or fifty.

BIBLIOGRAPHY Chandler, A.D. Jr (1977), The Visible Hand: The Managerial Revolution in American Business, Cambridge, MA, USA: Belknap Press of Harvard University Press. Cole, R. and R. Scott (2000), The Quality Movement and Organization Theory, Thousand Oaks, CA, USA: Sage Publications. Nelson, Richard R. and B. Sampat (2001), ‘Making sense of institutions as a factor in economic growth’, Journal of Economic Organization and Behavior, 44, 31–54. Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Belknap Press of Harvard University Press. Nelson, Richard R., Alexander Peterhansl and Bhaven Sampat (2004), ‘Why and how innovations get adopted: a tale of four models’, Industrial and Corporate Change, 13(5), 679–99. Prencipe, Andrea, Andrew Davies and Mike Hobday (eds) (2003), The Business of Systems Integration, Oxford, UK: Oxford University Press. Zollo, Maurizio and Sidney G. Winter (2002), ‘Deliberate learning and the evolution of dynamic capabilities’, Organizational Science, 13(3), 339–51.

3. The nature and replication of routines Geoﬀrey M. Hodgson The significance of routines within organizations is widely appreciated.1 Without routines, people within organizations would be overburdened with additional interpersonal negotiations and choices. Routines provide modularized sets of skills or capacities, involving relatively durable structured relations between individuals. Given their importance, it is necessary to understand both how they can be built and how they can be changed. Such awareness is essential to any analysis of how knowledge is retained and transferred, for the development of business strategy, and for the creation of policies to encourage more beneficial business practices. Detailed empirical investigation is vital in this regard, but detailed taxonomic studies based on empirical evidence are relatively rare. One reason why empirical investigations have so far remained limited and problematic is that the conceptual specification of a routine remains obscure. Greater conceptual precision is an essential precondition of fruitful empirical enquiry. This chapter attempts to illuminate the concept of the routine, by citing relevant insights from philosophy, social theory and psychology, and by focusing on some milestone contributions in this area. It is divided into five sections. The first section addresses the analogous and component concept of habit, with a view to making a distinction between habits and routines. The second section explores the metaphor of ‘routines as genes’ and argues that routines must be treated as capacities or dispositions, rather than behaviours. The third section considers how routines persist and carry information through time. The fourth section raises the issue of what is meant by the replication of routines, after briefly citing some theoretical and empirical studies in the area. The fifth section concludes the chapter.

1.

HABITS AS THE BASIS AND INDIVIDUAL ANALOGUE OF ROUTINES

The idea of a habit is elemental for understanding the concept of a routine. Routines operate through the triggering of individual habits and they are the organizational analogue of individual habits. 26

The nature and replication of routines

27

Like ‘routine’, the word ‘habit’ exists in common parlance and has a variety of meanings. Both words need to be defined more precisely for scientific usage. Even in scientific circles, and especially since the rise of behaviourist psychology after the First World War, there has been some ambiguity in the definition of habit. Influenced by positivist philosophy, behaviourists proposed the unworkable condition that all scientific entities had to be directly observable. Willard van Orman Quine and others (1953) showed that this position was untenable because all science is based on unobservable presuppositions. This anti-positivist argument is now widely accepted by philosophers of science. One of the negative outcomes of behaviourist psychology was the rejection of important ideas such as habit, or their reinterpretation in purely behavioural terms. The concept of habit was central to the pragmatist philosophy and institutional economics of the early twentieth century. It is useful to return to the meaning of the term employed then, especially as this earlier usage is now enjoying a revival. Pragmatist philosophers and institutional economists such as Thorstein Veblen (1919) regarded habit as an acquired proclivity or capacity, which may or may not be actually expressed in current behaviour (Hodgson, 2004). Repeated behaviour is important in establishing a habit. But habit and behaviour are not the same. If we acquire a habit we do not necessarily use it all the time. It is a propensity to behave in a particular way in a particular class of situations. The pragmatist philosopher and psychologist William James (1892: 143) wrote: ‘Habit is thus the enormous fly-wheel of society, its most precious conservative agent.’ Similarly, the pragmatist sociologists William Thomas and Florian Znaniecki (1920: 1851) criticized ‘the indistinct use of the term “habit” to indicate any uniformities of behavior. . . . A habit . . . is the tendency to repeat the same act in similar material conditions.’ Also treating habit as a propensity, William McDougall (1908: 37) wrote of ‘acquired habits of thought and action’ as ‘springs of action’ and saw ‘habit as a source of impulse or motive power’. Elsewhere, in his defences against the behaviourist invasion in psychology, McDougall (1924) explicitly emphasized the conceptual diﬀerence between dispositions and behaviour. As the pragmatist philosopher and psychologist John Dewey (1922: 42) put it: ‘The essence of habit is an acquired predisposition to ways or modes of response.’ The use of habit is largely unconscious. Habits are submerged repertoires of potential behaviour; they can be triggered or reinforced by an appropriate stimulus or context.2 The pragmatist philosopher Charles Sanders Peirce (1878: 294) declared, the ‘essence of belief is the establishment of habit’. Accordingly, habit is not the negation of deliberation, but its necessary foundation. Reasons and

28

Conceptual foundations

beliefs are often the rationalizations of deep-seated feelings and emotions that spring from habits that are laid down by repeated behaviours. This interplay of behaviour, habit, emotion and rationalization helps to explain the normative power of custom in human society. Hence ‘custom reconciles us to everything’ – as Edmund Burke wrote long ago – and customary rules can acquire the force of moral authority. In turn, these moral norms help to further reinforce the institution in question. Habits are socially acquired, not genetically transmitted. In this respect they are distinguished from instincts. Instincts are blunt instruments to deal with changing, complex and unpredictable circumstances. Humans developed the capacity to acquire habits concomitantly with the evolution of a cultural apparatus by which adaptive solutions to problems of survival could be preserved and passed on (Veblen, 1914: 6–7; Richerson and Boyd, 2001; Hodgson, 2004). By accepting the foundational role of habit in sustaining rule-following behaviour, we can begin to build an alternative ontology of institutions and routines, in which we avoid the conceptual problems of an account based primarily on intentionality. This is not to deny the importance of intentionality, but to regard it as a consequence as much as a cause, and to place it in the broader context of other, non-deliberative behaviours. Importantly, all learning, and the attainment of all skills, depends on the acquisition of habits. Knowledge and skills involve the capacity to address a complex problem and to identify rapidly the means of dealing with it. Experience and intuition are crucial here, and these must be grounded in acquired habits of thought or behaviour that dispose the agent to identify the crucial aspects of or responses to the problem. All skills, from knowledge of mathematics through competence with languages to ability with a musical instrument, depend on habits. Habits are the necessary means of avoiding full reflection over every detail, so that the more deliberative levels of the mind are freed up for more strategic issues. If all details were necessarily the subject of conscious deliberation, then the mind would be overwhelmed and paralysed by minutiae. We briefly consider two possible types of mechanism by which habit may be replicated from person to person.3 The first is by incentive or constraint. These can provide reasons to acquire specific customs, follow particular traﬃc conventions and use specific linguistic terms. In these cases, because others are acting in a particular way we can have powerful incentives to behave accordingly. In doing so, we too build up habits associated with these behaviours. The behaviours are reproduced and also the habits giving rise to them are replicated. Another possible mechanism is imitation. Imitation need not be fully conscious, and it will also involve some ‘tacit learning’ (Polanyi,

The nature and replication of routines

29

1967; Reber, 1993; Knudsen, 2002). Perhaps imitation can occur even without strong incentives, on the grounds that the propensity to imitate is instinctive, and this instinct has itself evolved for eﬃcacious reasons among social creatures (James, 1892; Veblen, 1899; Campbell, 1975; Boyd and Richerson, 1985; Simon, 1990; Tomasello, 2000; Henrich, 2004). However, an imitation instinct would require an existing set of common behaviours in the group, otherwise an emerging propensity to imitate might not have a selection advantage. For instinctive imitation to take oﬀ, common behaviours may have to emerge for other reasons. Furthermore, if imitation is more than mimicry, then the rules and understandings associated with it also have to be transmitted. Imitation is more problematic than it appears. Nevertheless, there are provisional grounds to consider a partially instinctive propensity to imitate as a strong element in the complex social glue, and hence a force behind the replication of habits. If habits replicate, then this means that multiple habits exist with similar characteristics. With genes, replicative similarity is at the level of the generative structure. However, there is no obvious reason why two people with similar habits have similar neural patterns in the brain. Habit similarity is at the level of descriptions of the rule-like dispositions that give rise to the behaviour, not of some neural structure. Having established the concept of habit, and for reasons that should become clearer below, we are now in a stronger position to turn to the concept of a routine. In the following section it will be explained how routines play a similar role for organizations that habits play for individuals.

2.

ROUTINES AS ORGANIZATIONAL GENES

In everyday parlance the word ‘routine’ is used loosely to refer to repeated sequences of behaviour, by individuals as well as by organizations. However, with their metaphor of ‘routines as genes’, Richard Nelson and Sidney Winter (1982) suggested a more specific and technical meaning for the term, which is further elaborated here. In evolutionary and institutional economics a consensus has emerged that routines relate to groups or organizations, whereas habits relate to individuals (Cohen et al., 1996; Dosi et al., 2000). Individuals have habits; groups have routines. Routines are the organizational analogue of habits. But routines do not simply refer to habits that are shared by many individuals in an organization or group. If this were the case there would be no need for the additional concept of a routine. Routines are not reducible to habits alone: they are organizational meta-habits, existing on a substrate of

30

Conceptual foundations

habituated individuals in a social structure. Routines are one ontological layer above habits themselves. This does not mean that routines exist independently of individuals or that individuals take a subsidiary place in the analysis. Without individuals there would be neither social structures nor routines. Routines exist because structured interactions of individuals give rise to emergent properties that (by definition) are not properties of individuals taken severally. Similarly, water has additional properties that are not possessed by separate hydrogen or oxygen atoms. The concept of an emergent property is widely adopted in the philosophy of science. It means that we cannot explain properties of wholes by focusing exclusively on the properties of parts, without considering structured causal interactions between those parts. This means that attempts to explain routines and other organizational phenomena in terms of individuals alone, without considering their structured interactions, are doomed.4 Having established the analogy between habits and routines, in the next section the causal connection between the two will be examined in more detail. It is first necessary to address an important question concerning the nature of routines: are they organizational dispositions or organizational behaviours? Confusingly, in their 1982 book, Nelson and Winter sometimes treat routines as dispositions, but otherwise described them as behaviours. For example, Nelson and Winter (1982: 15) write: ‘It is that most of what is regular and predictable about business behavior is plausibly subsumed under the heading “routine”’. But they go on in the same sentence to describe routines as ‘dispositions . . . that shape the approach of the firm’ to problems. Routines are also treated as ‘organizational memory’, which refers more to capabilities than to behaviour. Another passage introduces the useful analogy between a routine and a computer program, but repeats the same confusion. Nelson and Winter (1982: 97) see a ‘routine’ as being like a computer ‘program’, referring thereby ‘to a repetitive pattern of activity in an entire organization’ as well as to skills or capacities. But there is a diﬀerence between a computer program and the computer’s output or behaviour. The computer program is a rule-based system, with a generative coding that, along with other inputs, determines the computer’s output or behaviour. Nelson and Winter conflate generative and dispositional factors such as the computer program with outputs such as a ‘repetitive pattern of activity’ or ‘performance’. Nelson and Winter (1982) refer repeatedly to ‘routines as genes’. This is another useful analogy. But of course, routines are very diﬀerent from genes. Routines do not replicate biologically and they are much less enduring. All analogies are inexact in some respects and must be handled with

The nature and replication of routines

31

care, as Nelson and Winter are fully aware. The gene analogy usefully points to routines as relatively durable carriers of information through shorter periods of time, with the algorithmic capacity to generate particular outcomes in given circumstances. Routines are like genes in the sense that they are both generative, rule-like structures and potentialities. However, routines (like genes) cannot be both generative structures and outcomes of such structures. This point is not about the appropriateness or otherwise of biological analogies, but about the clear meanings of words and their ontological references. Winter (1995: 149–50) distinguishes between a ‘routine in operation at a particular site . . . a web of coordinating relationships connecting specific resources’ and the ‘routine per se – the abstract activity pattern’. But the one term ‘routine’ cannot apply to both the ‘web of coordinating relationships’ and the ‘activity pattern’ that is the outcome of the coordinating structure and its environmental triggers; it cannot usefully denote both potentiality and actuality. It has to denote one or the other, but not both. At root there is a philosophical contention here: the essence of what an entity is cannot be entirely appraised in terms of what an entity does. If we make this confusion, then we wrongly imply that when the entity interrupts its characteristic activity, then it ceases to exist. Birds fly. But what defines a bird is the (existing or past) capacity to fly, not flying itself. If a bird were wrongly defined as a flying animal, then any bird sitting on a branch or pecking on the ground would cease to be a bird. Accordingly routines are not behaviour; they are stored behavioural capacities or capabilities. Consider a firm in which all employees and managers work between 9a.m. and 5p.m. only. During this working day a number of organizational routines can be energized. At other times the firm is inactive. But the routines do not all disappear at 5p.m., to reappear mysteriously the next day. The routines-as-capacities remain, as long as the individuals have the capacity and disposition to work again together in the same context. Subject to this condition, the routines can be triggered the next day by appropriate stimuli. Aristotle made the central philosophical point here, more than 2300 years ago. In his Metaphysics Aristotle (1956: 227–8) pointed out that individual skills are retained for a while even if they are unused. Otherwise how would work using the skill be resumed? This is an obvious – even trivial – point, but it has implications for those who define routines in terms of behaviour. Aristotle insisted that ‘we must obviously draw a distinction between potentiality and actuality’ and argued that the essence of an entity lies in its potentiality, rather than the observable outcomes. Another enduringly relevant point here is that definitions or ontologies that are based on behaviour cannot cope with instances where the

32

Conceptual foundations

behaviour changes or ceases. Even when an entity with a particular behavioural propensity is inert, the capacity to produce the characteristic behaviour remains, and this capacity, not the outcome, continues to define part of the essence of the entity. Although ancient, this point is not arcane; it is widely utilized in modern realist philosophy of science. Central to most strands of modern realist philosophy is the distinction between the potential and the actual, between dispositions and outcomes, where in each case the former are more fundamental than the latter. 5 Science is about the discovery of causal laws or principles. Causes are not events; they are generative mechanisms that can under specific conditions give rise to specific events. For example, a force impinging on an object does not always make that object move. The outcome also depends on friction, countervailing forces, and other factors. Causes create potentialities; they are not necessarily realized in outcomes. As Veblen (1919: 89) put it: ‘The laws of nature are . . . of the nature of a propensity.’ Hence there must be a distinction between an observed empirical regularity and any causal law that lies behind it. Similarly there must be a distinction between the capacities and behaviours of an entity. In biology, genes and genotypes are potentialities; they are not behaviours. In the socio-economic domain, the closest analogies to genotypes are the generative rule-like structures inherent in ingrained individual habits and in organizational routines. Habits and routines are thus understood as conditional, rule-like potentialities or dispositions, rather than behaviour as such. The key distinction in the socio-economic sphere is between habits and routines as dispositions, on the one hand, and manifest behaviour, on the other hand. Similarly, it is appropriate to treat routines as behavioural propensities of organizations, rather than behaviour as such. In this light, any emphasis on the allegedly predictable character of routines is misplaced. Predictions relate to outcomes or events, not to causal laws, rules or generative structures. The moderately dependable feature of a routine, rule or computer program is not one of predictability but of durability. Routines (or rules or computer programs) are usually conditional on other inputs or events. As a result any predictability does not stem from the routine alone but from the predictability of these other inputs. For example, a firm may have a fixed mark-up pricing routine of adding 20 per cent to the unit cost of its products. If costs were capricious and highly variable, as they might be under some circumstances, then the resulting price would be equally unreliable. The relatively enduring and persistent quality of a routine is not its outcome but its generative, rule-like structure. While a consensus has been established that a routine is an organizational rather than an individual phenomenon, some confusion remains on

The nature and replication of routines

33

the above points, and this has led to some conceptual and empirical diﬃculties.6 Some of these diﬃculties can be overcome by consistently treating a routine as an organizational capacity and generative structure, analogous to biological genes or computer programs, but having distinctive features of their own. To their credit, both Nelson and Winter are now more inclined to describe the routine in terms of a capacity. Nelson and Winter (2002: 30) write: ‘we treat organizational routine as the organizational analogue of individual skill.’ A similar attitude is evident elsewhere. Barbara Levitt and James March (1988: 320) write: ‘The generic term “routines” includes the forms, rules, procedures, conventions, strategies, and technologies around which organizations are constructed and through which they operate.’ Another useful definition of a routine as a potentiality or capability, rather than behaviour, is found in the discussion in Michael Cohen et al. (1996: 683), ‘A routine is an executable capability for repeated performance in some context that [has] been learned by an organization in response to selective pressures.’ A routine is here defined as a generative structure or capacity within an organization. Routines are organizational dispositions to energize conditional patterns of behaviour within an organized group of individuals, involving sequential responses to cues. The next section raises the general questions of how routines work within organizations and how they carry information.

3.

HOW DO ROUTINES CARRY INFORMATION?

The analysis of how routines endure and replicate is enormous and incomplete (Hodgson, 2003). At present, our general understanding is limited, and progress depends largely on the accumulation of detailed case studies. As Winter (1990: 270) notes, so far ‘little attention has been paid to the mechanism by which whatever-it-is-called is transmitted’ and to its ‘replication mechanism’. For Winter (1990: 294 n.) this amounts to a regrettable ‘vagueness on a key issue’. As Winter (1990: 270–75) insisted: ‘The question of what is “inherited” and how the inheritance mechanisms works is, however, . . . central and . . . far from definitive resolution . . . To develop the routines as genes approach fully, the problem of inheritance mechanisms needs to be dealt with convincingly.’ To understand how routines work it is necessary to consider how any tacit or other information associated with a routine is preserved and replicated. A very useful study in this regard is by Michael Cohen and Paul Bacdayan (1994). They use the distinction in psychology between

34

Conceptual foundations

procedural and other, more cognitive forms of memory, such as semantic, episodic or declarative memory. As psychologists Endel Tulving and Daniel Schacter (1990: 301) put it: ‘The domain of procedural memory is behavior, whereas that of semantic and episodic memory is cognition or thought. Cognitive memory systems have the capability of modeling the external world – that is, of storing representations of objects, events, and relations among them – whereas procedural memory does not have this capacity.’ Procedural memory is triggered by preceding events and stimuli. It typically leads to behavioural responses and has a major tacit component. It is potential action that is energized by social or other cues, often involving physical artefacts. ‘Procedural knowledge is less subject to decay, less explicitly accessible, and less easy to transfer to novel circumstances’ (Cohen and Bacdayan, 1994: 557). Routines depend upon a structured group of individuals, each with habits of a particular kind, where many of these habits depend upon procedural memory. The behavioural cues by some members of a structured assembly of habituated individuals triggers specific habits in others. Hence various individual habits sustain each other in an interlocking structure of reciprocating individual behaviours. Together these behaviours take on collective qualities associated with teams. But both individuals and structures are involved throughout. The organization or group provides a structured social and physical environment for each individual, including rules and norms of behaviour, of both the explicit and the informal kind. This environment is made up of the other individuals, the relations between them and the technological and physical artefacts that they may use in their interactions. This social and physical environment enables, stimulates and channels individual activities, which in turn can help trigger the behaviour of others, produce or modify some artefacts, and help to change or replicate parts of this social and physical environment. Partly because of procedural memory, organizations can have important additional properties and capacities that are not possessed by individuals, taken severally. The organization provides the social and physical environment that is necessary to enable specific activities, cue individual habits and deploy individual memories. If one person leaves the organization and is replaced by another, then the new recruit may have to learn the habits that are required to maintain specific routines. Just as the human body has a life in addition to its constituent cells, the organization thus has a life in addition to its members. Generally, the organizational whole is greater than the sum of the properties of its individual members, taken severally. The additional properties of the whole stem from the structured relations and causal interactions between the individuals involved. This

The nature and replication of routines

35

is a central proposition in the emergentist tradition of philosophy and social theory (Blitz, 1992; Kontopoulos, 1993; Hodgson, 2004; Weissman, 2000). A routine derives from the capacity of an organization to provide conditions to energize a series of conditional, interlocking, sequential behaviours among several individuals within the organization. Cohen and Bacdayan (1994: 557) write: ‘The routine of a group can be viewed as the concatenation of such procedurally stored actions, each primed by and priming the actions of others.’ This statement captures the dependence of routines on procedural memory, but is somewhat ambiguous concerning the genotypic or phenotypic status of a routine. As argued above, routines are not behaviour; they are stored behavioural capacities or capabilities. These capacities involve knowledge and memory. They involve organizational structures and individual habits which, when triggered, lead to sequential behaviours. But this does not mean that a routine can be fully codified. Routines are not necessarily nominal, codified or oﬃcially approved procedures. Routines generally rely on informal and tacit knowledge, and this fact is clearly relevant for understanding their replication. The temporal durability of routines and the way that they can embody knowledge ‘forgotten’ by individuals is illustrated by an anecdote related by Elting Morison (1966). A time-and-motion expert was studying film footage of Second World War motorized artillery crews. He was puzzled by a recurring three-second pause just before the guns were fired. An old soldier also watching the film suddenly realized that the three-second pause had originated from the earlier era in which the guns were drawn by horses, and the horses had to be held and calmed in the seconds just before the guns went oﬀ. Despite its eventual redundancy, this part of the routine had survived the transition from horse-driven to motorized artillery. Part of the knowledge held in a routine can become obsolete, yet still be reproduced, like the examples of ‘rudimentary organs’ discussed by Charles Darwin (1859: 450–58). In sum, the explanation of the persistence of routines is a story of both the persistence of organizations and the persistence of the habits of the individuals involved. The explanation of the latter is in large part a psychological matter, referring to the nature of individual habits. By contrast, explanations of the former must answer the question of why organizations are often resistant to change, which has been an important question within organization theory since its inception (Simon, 1947; March and Simon, 1958; Stinchcombe, 1965). There is no attempt to develop such an account here: it is simply identified as a necessary component of the explanation of the persistence of routines.

Conceptual foundations

36

4.

THE REPLICATION OF ROUTINES

Just as habits replicate from individual to individual, routines replicate from group to group and from organization to organization. In studies of technological diﬀusion, organization studies and the strategic management literature there is some discussion of the diﬀusion or replication of routines (Aldrich and Martinez, 2003; Becker and Lazaric, 2003; DiMaggio and Powell, 1983; Hannan and Freeman, 1984, 1989; Lazaric and Denis, 2001; Levitt and March, 1988; Rogers, 1995; Stinchcombe, 1990; Szulanski 1996, 2000; Zucker, 1987). Prominent mechanisms for the diﬀusion of routines involve the movement of employees from organization to organization, or independent experts or consultants that help to transfer knowledge and experience gained in one context to another. The above authors cite case studies involving the transfer of technologies, management procedures, corporate multidivisional structures, accounting conventions and much else. What is central to these transfers is the replication of practices and organizational relationships. What is generally critical is the capacity of the receiving organization to accommodate and utilize these practices and relationships in the context of its own ingrained culture of habits and beliefs. Are these examples of routine replication, analogous to the replication of genes in biology? To answer this question we need to employ the general distinction between a ‘replicator’ and an ‘interactor’, as found in the philosophy of evolutionary systems (Hull, 1988). The concept of replication, as developed by Kim Sterelny et al. (1996), Peter Godfrey-Smith (2000), Dan Sperber (2000) and Robert Aunger (2002), involves the following three conditions: ●

● ●

Causation: the source must be causally involved in the production of the copy, at least in the sense that without the source the particular copy would not be created. Similarity: the copy must also possess the capacity to replicate and be like its source in other relevant respects. Information transfer: the process that generates the copy must obtain the information that makes the copy similar to its source from that same source.

Sperber (2000) argues that the third condition is critical. Cases where the third condition is absent include the diﬀusion of behaviour, such as contagious laughter, where A triggers B, but where the propensity to that behaviour is already present in B. Sperber argues convincingly that many examples of so-called ‘memes’ do not satisfy the third condition and thus are not true replicators.

The nature and replication of routines

37

Relevant ‘contagious’ behaviour in the socio-economic context would be the spread of trustworthy/untrustworthy, civil/uncivil, or cooperative/ uncooperative behaviours. These diﬀusions of behaviour are triggered by external stimuli, but generally the behaviour is already latent in the individual involved. Such diﬀusions of similar behaviour do not obtain their characteristics of similarity from the stimulus that each receives. For true replication to occur, the capacity to produce the behaviour must itself be replicated. Again this points to the importance of making a distinction between capacity or disposition, on the one hand, and behaviour on the other. In addition, more must be involved than the replication of the individual habits associated with the routine. The replication of routines must involve the replication of the generative structures and capacities that are additional to the habits of the individuals involved, taken severally. Furthermore, given that the structures associated with the routine are constitutive of the routine itself, and contain some of the ‘information’ embodied in the routine, these structures must play a causal role in a true process of replication. The transfer of skills from a master to an apprentice is typically a case of habit replication, rather than the replication of routines. For routine replication to occur, not only must the individual skills be replicated, but also the manner in which they are organized together into an eﬀective structured relationship between several individuals. The replication of individual skills and habits involves the transfer of tacit as well as codifiable knowledge. Tacit knowledge is transferred as a result of repeated practice, often with similar stimuli and constraints. Because routines are partly made up of individual skills, tacit knowledge is unavoidably involved at the level of routines as well (Polanyi, 1967; Nelson and Winter, 1982; Hannan and Freeman, 1989; Cohen and Bacdayan, 1994; Cohen et al., 1996). This tacit dimension, combined with its structured nature of a routine, makes it a complex and elusive entity, which is often diﬃcult to replicate. As with habits, replicative similarity is at the level of descriptions of the rule-like dispositions that give rise to the behaviour, rather than at the level of the neural patterns of the individuals involved.7 The facilitation of the process of routine replication is a key area for management practice and enquiry. It may be possible to replicate some routines through descriptions of their essentials in a manual or textbook, but these cases are relatively simple and rare, and will often rely on subsequent practice and advice. In the organization studies literature, examples of successful routine replication typically involve the combination of codifiable information and instructions with extensive personal example, advice and contact, where

38

Conceptual foundations

the receiving organization has suﬃcient plasticity to usefully absorb and accommodate the routine. Sometimes routines are spread as a result of laws or rules that emanate from a third organization, such as the state, or an association of employers. Otherwise the replication of routines can occur as the result of the strategy of its receiving organization, or it can result from lower-level contact, stimulation and imitation. Routines replicate, and they do so on a substrate of organized and habituated individuals. Another question that arises concerns the propensities of organizations to replicate their routines. At the individual level, habits replicate largely because individuals have an instinctive propensity to imitate some behaviours. This involves the ability to sense the more significant actions, and the tacit rules and meanings associated with behaviour (Tomasello, 2000). There is no equivalent to this individual instinct at the level of the organization. On the contrary, firms are often conservative and diﬃcult to change. The structure and culture of the organization often discourages change, as this would challenge individual vested interests and the ‘groupthink’ of the collective. Rather than imitating others, organizations will often carry on in the same old way, unless their internal perception of a crisis prompts an energetic survival strategy. Complex organizations are extremely recent in human history, so despite the possible survival advantages of organizational imitation, the evolution of selective capacities to imitate at the organizational level has had little time to develop. Light has been shown on the degree of organizational conservatism by empirical enquiry. Michael Hannan and John Freeman (1989) are leading proponents of the view that the capacity to change routines within organizations is relatively limited, and that changes in the population of routines within industries or societies largely comes about through the survival or extinction of specific organizations, and the consequent persistence or disappearance of the routines they carry, rather than through modifications in the routines themselves. This is an important area of ongoing empirical enquiry.8 What then are the corresponding interactors? David Hull (1988: 408) defines an interactor as ‘an entity that directly interacts as a cohesive whole with its environment in such a way that this interaction causes replication to be diﬀerential’. The term ‘cohesive whole’ indicates that its components stick together and remain united. This must mean at least that all the components depend critically on the survival of the whole, and that to some degree the components depend on the survival of each other. Refining this definition still further, Geoﬀrey Hodgson and Thorbjørn Knudsen (2004) argue that a firm may be regarded as an interactor, and consequently as a ‘vehicle’ for its inherent habits and routines. The fate of a routine is often dependent on the fate of its host firm. It should be pointed out, however,

The nature and replication of routines

39

that although this type of evolutionary approach has a long history (Hodgson, 2004), at least in its present form and context it is in the early stages of its development, and many outstanding conceptual problems remain to be resolved.

5.

CONCLUSION

This chapter has explored the concept of a routine in fundamental terms, using insights from philosophy, psychology and social theory. A routine is here defined as a generative structure or capacity within an organization. By definition, routines are organizational dispositions to energize conditional patterns of behaviour within an organized group of individuals, involving sequential responses to cues. There are important philosophical reasons, endorsed by modern philosophy of science, why routines should be defined as organizational dispositions or capacities, rather than behaviour as such. Just as habits relate to individuals, routines relate to organizations. Both are socially transmitted dispositions, formed through repeated behaviours. Routines themselves are structures of interlocking individual habits. But routines are more than mere aggregations of habits, because they also depend on the emergent properties of organization itself, emanating from structured causal relations and interactions between individuals. One of the reasons why the study of routines is important for the study of business practice is that they are repositories and carriers of knowledge and skill. The routine is often the means through which individual skills are triggered and energized. One psychological mechanism that is important here is procedural memory, which means that some powers of recall can be enhanced when triggered by cues provided by others. In this manner the routine as a whole becomes more than the sum of the capacities of the individuals involved, taken severally. Hopefully, these arguments demonstrate that the philosophy of science can be useful in refining key concepts such as a routine. This literature helps us understand the fact that routines have properties that are not possessed by individuals themselves. Furthermore, the philosophy of biology provides general concepts such as replication that can also be applied to routines (Hodgson and Knudsen, 2006a). The study of routines is part of a broader project to apply evolutionary principles to the analysis of social and economic development. This project must involve an alliance of conceptual development, formal theory and empirical enquiry. It is unlikely that significant overall progress will be made without advancement on each of these three fronts. Just as philosophical

40

Conceptual foundations

discussion cannot give us all the answers without some empirical application, empirical research on routines and other socio-economic phenomena requires sharpened conceptual and theoretical tools.

NOTES 1. Thanks are due to participants at the Second Routines Workshop in Sophia-Antipolis, France, 21–22 January 2005 for comments on a previous version of this paper. Use is made of material from Hodgson (2008). 2. The conception of a habit as a propensity or disposition is also found in modern works such as Camic (1986), Margolis (1994), Murphy (1994), and Kilpinen (2000). Contemporary psychologists also adopt a similar concept of habit: Ouellette and Wood (1998), Wood et al. (2002). 3. See the more extensive discussions in Knudsen (2008). 4. See, for example, Bunge (1973–1989; 2000), Bhaskar (1975), Blitz (1992), Kontopoulos (1993), Weissman (2000), Hodgson (2004; 2007). An alternative concept within the philosophy of science is supervenience (Kim, 1993). In defending emergence, Humphreys (1997) points out that without it there would be no reason for the pursuit of other sciences, except subatomic physics. 5. For realist accounts upholding a distinction between generative mechanisms or causal powers, on the one hand, and outcomes or events, on the other, see for example Bhaskar (1975), Harré and Madden (1975), Popper (1990). 6. For discussions of some of these diﬃculties see Cohen et al. (1996), Reynaud (2000), Becker (2001, 2005) and Lazaric (2000). 7. It is possible that two routines with very similar sets of rules may lead to diﬀerent behavioural outcomes, because of diﬀerent information inputs or stimuli. Obversely, behaviourally similar routines may emanate from very diﬀerent rule structures. Related possibilities exist in biology with genes and behaviour. In all cases, with genes, habits and routines, relevant similarity exists at the level of description of conditional, rule-like dispositions. 8. Usher and Evans (1996) provide a useful review of this literature, with further evidence. However, their characterization of this debate as between ‘Darwinian’ and ‘Lamarckian’ concepts of change is at best highly misleading (Hodgson and Knudsen, 2006b).

REFERENCES Aldrich, Howard E. and Martha Martinez (2003), ‘Entrepreneurship as social construction: a multi-level evolutionary approach’, in Z.C. Acs and David B. Audretsch (eds), Handbook of Entrepreneurial Research, Boston, MA, USA: Kluwer, pp. 359–99. Aristotle (1956), Metaphysics, edited and translated by John Warrington with an Introduction by W. David Ross, London, UK: Dent. Aunger, Robert (2002), The Electric Meme: A New Theory of How We Think, New York, USA: Free Press. Becker, Markus C. (2001), ‘The role of routines in organisations: an empirical and taxonomic investigation’, PhD thesis, University of Cambridge, UK. Becker, Markus C. (2005), ‘The concept of routines: some clarifications’, Cambridge Journal of Economics, 29(2), March, 249–62.

The nature and replication of routines

41

Becker, Markus C. and Nathalie Lazaric (2003), ‘The influence of knowledge in the replication of routines’, Économie Appliquée, 56(3), September, 65–94. Bhaskar, Roy (1975), A Realist Theory of Science, 1st edn, Leeds, UK: Leeds Books. Blitz, David (1992), Emergent Evolution: Qualitative Novelty and the Levels of Reality, Dordrecht, Holland: Kluwer. Boyd, Robert and Peter J. Richerson (1985), Culture and the Evolutionary Process, Chicago, IL, USA: University of Chicago Press. Bunge, Mario A. (1973–1989), Treatise on Basic Philosophy, 8 vols, Dordrecht, Holland: Reidel. Bunge, Mario A. (2000), ‘Ten modes of individualism – none of which works – and their alternatives’, Philosophy of the Social Sciences, 30(3), 384–406. Camic, Charles (1986), ‘The matter of habit’, American Journal of Sociology, 91(5), March, 1039–87. Campbell, Donald T. (1975), ‘On the conflicts between biological and social evolution and between psychology and moral tradition’, American Psychologist, 30(12), December, 1103–26. Cohen, Michael D. and Paul Bacdayan (1994), ‘Organizational routines are stored as procedural memory – evidence from a laboratory study’, Organization Science, 5(4), November, 554–68. Cohen, Michael D., Roger Burkhart, Giovanni Dosi, Massimo Egidi, Luigi Marengo, Massimo Warglien and Sidney G. Winter (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5(3), 653–98. Darwin, Charles R. (1859), On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life, 1st edn, London, UK: Murray. Dewey, John (1922), Human Nature and Conduct: An Introduction to Social Psychology, 1st edn, New York, USA: Holt. DiMaggio, Paul J. and Walter W. Powell (1983), ‘The iron cage revisited: institutional isomorphism and collective rationality in organizational fields’, American Sociological Review, 48(2), April, 147–60. Dosi, Giovanni, Richard R. Nelson and Sidney G. Winter (2000), ‘Introduction: the nature and dynamics of organizational capabilities’, in Giovanni Dosi, Richard R. Nelson and Sidney G. Winter (eds), The Nature and Dynamics of Organizational Capabilities, Oxford, UK: Oxford University Press, pp. 1–22. Godfrey-Smith, Peter (2000), ‘The replicator in retrospect’, Biology and Philosophy, 15, 403–23. Hannan, Michael T. and John Freeman (1984), ‘Structural inertia and organizational change’, American Sociological Review, 49(2), April, 149–64. Hannan, Michael T. and John Freeman (1989), Organizational Ecology, Cambridge, MA, USA: Harvard University Press. Harré, Rom and Edward H. Madden (1975), Causal Powers: A Theory of Natural Necessity, Oxford, UK: Basil Blackwell. Henrich, Joseph (2004), ‘Cultural group selection, coevolutionary processes and large-scale cooperation’, Journal of Economic Behavior and Organization, 53(1), February, 3–35. Hodgson, Geoﬀrey M. (2003), ‘The mystery of the routine: the Darwinian destiny of An Evolutionary Theory of Economic Change’, Revue Économique, 54(2), March, 355–84.

42

Conceptual foundations

Hodgson, Geoﬀrey M. (2004), The Evolution of Institutional Economics: Agency, Structure and Darwinism in American Institutionalism, London, UK and New York, USA: Routledge. Hodgson, Geoﬀrey M. (2007), ‘Institutions and individuals: interaction and evolution’, Organization Studies, 28(1), 95–116. Hodgson, Geoﬀrey M. (2008), ‘The concept of a routine’, in Markus C. Becker (ed.), Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 15–28. Hodgson, Geoﬀrey M. and Thorbjørn Knudsen (2004), ‘The firm as an interactor: firms as vehicles for habits and routines’, Journal of Evolutionary Economics, 14(3), July, 281–307. Hodgson, Geoﬀrey M. and Thorbjørn Knudsen (2006a), ‘Why we need a generalized Darwinism: and why a generalized Darwinism is not enough’, Journal of Economic Behavior and Organization, 61(1), September, 1–19. Hodgson, Geoﬀrey M. and Thorbjørn Knudsen (2006b), ‘Dismantling Lamarckism: why descriptions of socio-economic evolution as Lamarckian are misleading’, Journal of Evolutionary Economics, 16(4), October, 343–66. Hull, David L. (1988), Science as a Process: An Evolutionary Account of the Social and Conceptual Development of Science, Chicago, IL, USA: University of Chicago Press. Humphreys, Paul (1997), ‘How properties emerge’, Philosophy of Science, 64(1), March, 1–17. James, William (1892), Psychology: Briefer Course, New York, USA and London, UK: Holt and Macmillan. Kilpinen, Erkki (2000), The Enormous Fly-Wheel of Society: Pragmatism’s Habitual Conception of Action and Social Theory, Helsinki, Finland: University of Helsinki. Kim, Jaegwon (1993), Supervenience and Mind, Cambridge, UK and New York, USA: Cambridge University Press. Knudsen, Thorbjørn (2002), ‘The significance of tacit knowledge in the evolution of human language’, Selection, 3(1), 93–112. Knudsen, Thorbjørn (2008), ‘Organizational routines in evolutionary theory’, in Markus C. Becker (ed.) Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 125–51. Kontopoulos, Kyriakos M. (1993), The Logics of Social Structure, Cambridge, UK: Cambridge University Press. Lazaric, Nathalie (2000), ‘The role of routines, rules and habits in collective learning: some epistemological and ontological considerations’, European Journal of Economic and Social Systems, 14(2), 157–71. Lazaric, Nathalie and Blandine Denis (2001), ‘How and why routines change: some lessons from the articulation of knowledge with ISO 9002 implementation in the food industry’, Économies et Sociétés, Série Dynamique technoloqique et organisation, 6(4), 585–611. Levitt, Barbara and James G. March (1988), ‘Organizational learning’, Annual Review of Sociology, 14, 319–40. March, James G. and Herbert A. Simon (1958), Organizations, New York, USA: Wiley. Margolis, Howard (1994), Paradigms and Barriers: How Habits of Mind Govern Scientific Beliefs, Chicago, IL, USA: University of Chicago Press. McDougall, William (1908), An Introduction to Social Psychology, 1st edn, London, UK: Methuen.

The nature and replication of routines

43

McDougall, William (1924), ‘Can sociology and social psychology dispense with instincts?’, American Journal of Sociology, 29(6), May, 657–73. Morison, Elting E. (1966), Men, Machines and Modern Times, Cambridge, MA, USA: MIT Press. Murphy, James Bernard (1994), ‘The kinds of order in society’, in Philip Mirowski (ed.), Natural Images in Economic Thought: “Markets Read in Tooth and Claw”, Cambridge, UK and New York, USA: Cambridge University Press, pp. 536–82. Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Harvard University Press. Nelson, Richard R. and Sidney G. Winter (2002), ‘Evolutionary theorizing in economics’, Journal of Economic Perspectives, 16(2), Spring, 23–46. Ouellette, Judith A. and Wendy Wood (1998), ‘Habit and intention in everyday life: the multiple processes by which past behavior predicts future behavior’, Psychological Bulletin, 124, 54–74. Peirce, Charles Sanders (1878), ‘How to make our ideas clear’, Popular Science Monthly, 12, January, 286–302. Polanyi, Michael (1967), The Tacit Dimension, London, UK: Routledge and Kegan Paul. Popper, Karl R. (1990), A World of Propensities, Bristol, UK: Thoemmes. Quine, Willard van Orman (1953), From a Logical Point of View, Cambridge, MA, USA: Harvard University Press. Reber, Arthur S. (1993), Implicit Learning and Tacit Knowledge: An Essay on the Cognitive Unconscious, Oxford, UK and New York, USA: Oxford University Press. Reynauld, Bénédicte (2000), ‘The properties of routines: tools of decision making and modes of coordination’, in Pier Paolo Saviotti and Bart Nooteboom (eds) (2005), Technology and Knowledge: From the Firm to Innovation Systems, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 249–62. Richerson, Peter J. and Robert Boyd (2001), ‘Built for speed, not for comfort: Darwinian theory and human culture’, History and Philosophy of the Life Sciences, 23(3/4), 423–63. Rogers, Everett M. (1995), Diﬀusion of Innovations, 3rd edn, New York, USA: Free Press. Simon, Herbert A. (1947), Administrative Behavior: A Study of Decision-Making Processes in Administrative Organization, 1st edn, New York, USA: Free Press. Simon, Herbert A. (1990), ‘A mechanism for social selection and successful altruism’, Science, 250, 21 December, 1665–8. Sperber, Dan (2000), ‘An objection to the memetic approach to culture’, in Robert Aunger (ed.), Darwinizing Culture: The Status of Memetics as a Science, Oxford, UK and New York, USA: Oxford University Press, pp. 162–73. Sterelny, Kim, Kelly C. Smith and Michael Dickison (1996), ‘The extended replicator’, Biology and Philosophy, 11, 377–403. Stinchcombe, Arthur L. (1965), ‘Social structure and organizations’, in James G. March (ed.), Handbook of Organizations, Chicago, IL, USA: Rand McNally, pp. 142–93. Stinchcombe, Arthur L. (1990), Information and Organizations, Berkeley, CA, USA: University of California Press. Szulanski, Gabriel (1996), ‘Exploring internal stickiness: impediments to the transfer of best practice within the firm’, Strategic Management Journal, 17, Winter Special Issue, pp. 27–43.

44

Conceptual foundations

Szulanski, Gabriel (2000), ‘Appropriability and the challenge of scope: Banc One routinizes replication’, in Giovanni Dosi, Richard R. Nelson and Sidney G. Winter (eds), The Nature and Dynamics of Organizational Capabilities, Oxford, UK: Oxford University Press, pp. 69–98. Thomas, William and Florian Znaniecki (1920), The Polish Peasant in Europe and America, vol. 2, New York, USA: Octagon. Tomasello, Michael (2000), The Cultural Origins of Human Cognition, Cambridge, MA, USA: Harvard University Press. Tulving, Endel and Daniel L. Schacter (1990), ‘Priming and human memory systems’, Science, 247(4940), 19 January, 301–6. Usher, John M. and Martin G. Evans (1996), ‘Life and death along gasoline alley: Darwinian and Lamarckian processes in a diﬀerentiating population’, Academy of Management Journal, 39(5), October, 1428–66. Veblen, Thorstein B. (1899), The Theory of the Leisure Class: An Economic Study in the Evolution of Institutions, New York, USA: Macmillan. Veblen, Thorstein B. (1914), The Instinct of Workmanship, and the State of the Industrial Arts, New York, USA: Macmillan. Veblen, Thorstein B. (1919), The Place of Science in Modern Civilization and Other Essays, New York, USA: Huebsch. Weissman, David (2000), A Social Ontology, New Haven, CT, USA: Yale University Press. Winter, Sidney G., Jr (1990), ‘Survival, selection, and inheritance in evolutionary theories of organization’, in Jitendra V. Singh (ed.), Organizational Evolution: New Directions, London, UK: Sage, pp. 269–97. Winter, Sidney G., Jr (1995), ‘Four Rs of profitability: rents, resources, routines, and replication’, in Cynthia A. Montgomery (ed.), Resource-Based and Evolutionary Theories of the Firm: Towards a Synthesis, Boston, MA, USA: Kluwer, pp. 147–78. Wood, Wendy, Jeﬀrey M. Quinn and D. Kashy (2002), ‘Habits in everyday life: thought, emotion, and action’, Journal of Personality and Social Psychology, 83, 1281–97. Zucker, Lynne G. (1987), ‘Institutional theories of organization’, Annual Review of Sociology, 13, 443–64.

PART II

Methods for analysing organizational routines

4.

Using workflow data to explore the structure of an organizational routine Brian T. Pentland, Thorvald Haerem and Derek W. Hillison

INTRODUCTION Empirical study of organizational routines poses many diﬃculties for the researcher. Routines are typically distributed in time, space and throughout an organization’s structure. Short of ‘stapling yourself to the paperwork’, it is diﬃcult to observe even a single performance of an organizational routine from beginning to end. In addition, the natural variability in performances can make it diﬃcult to identify a single representative pattern. Time, money and patience often limit us to observations of a few performances, or parts of performances or interviews with a subset of participants. Workflow systems provide an unprecedented opportunity to gather data about the patterns of action generated by routines (van der Aalst et al., 2003). With the proliferation of computer network technology, more and more organizations have adopted workflow systems to support their routines (Basu and Kumar, 2002). These systems typically involve a mixture of human and automatic processing – they are like the ‘glue’ that holds together other, more common applications (Becker et al., 2002). Typical workflow systems generate ‘event logs’ that include time-stamped records of each event or action that occurs in the system, making it possible to collect large numbers of performances at very low cost. Computer scientists have made tremendous progress in analysing event logs for a range of purposes, such as the recovery of formal process models (van der Aalst et al., 2003; van der Aalst and Weijters, 2004). Organizational scholars, however, have not paid as much attention to this increasingly ubiquitous source of data about organizational processes and routines. In this chapter, we demonstrate how workflow event logs can be used to analyse organizational routines. We begin by discussing the challenges 47

48

Methods for analysing organizational routines

involved in analysing routines as systems that generate patterns of events that can change over time. We provide a survey of techniques for analysing and representing the structure of organizational routines and illustrate these techniques using over 2000 performances of an invoice approval routine, as captured in a workflow management system. While workflow event logs have several important limitations, it seems clear that they provide a valuable source of data about organizational routines.

THEORY To have an empirical science of organizational routines, we need to be able to identify particular routines and compare them to other routines. These analytical moves – identification and comparison – are fundamental to any empirical science. In disciplines where individuals or organizations are the unit of analysis, such as psychology or economics, identification and comparison are easier. The additional complexity of studying patterns of behaviours by groups of individuals that take place in the context of an organization adds the diﬃculty of arbitrary boundaries and cases that span ascribed boundaries between units. Identification and comparison are particularly challenging if we conceptualize routines as generative systems (Feldman and Pentland, 2003; Pentland and Feldman, 2005). We are able to observe surface structure (the performative aspects of the routine), but we would like to say something about the underlying structures which generate the performances. The grammatical metaphor of ‘surface’ vs. ‘deep’ structure has limitations when applied to social systems (Pentland, 1995), but the basic analogy is useful when trying to understand human perceptions of complex tasks and systems (Chi et al., 1981; Haerem and Rau, 2007): we are trying to understand the features of a system that can generate a potentially infinite variety of performances from a finite sample of performances. When we say that organizational routines are generative systems, it means that there is some underlying mechanism that generates the interdependent patterns of action that we recognize as an organizational routine. Figure 4.1 illustrates this perspective. Figure 4.1 assumes that there is one routine generating the performances we are seeing. But in field research, the situation is more like Figure 4.2: we see a collection of performances and we would like to know something about the underlying routine. Is there one routine, or many? Which performances represent standard practice and which are exceptions? More broadly, how can we generalize about a routine from observing a sample of its performances?

Using workflow data to explore the structure of an organizational routine Routine

Generates

Performances A

B A

Ostensive Pattern A B C D

49

C C

C

D D

D

(etc)

Figure 4.1

Routine as generative structure Performances

Indicate

Routine?

A B C D A C D C D

?

(etc)

Figure 4.2

Generalization of structure from specific performances

We can see this problem in two ways: measurement or induction. As a measurement problem, we are using the performances to compute various properties of the routine, such as the average cycle time or the number of steps. Because each performance has a beginning and an end, we can compute the cycle time. The variability of the performances is easily dealt with by familiar procedures such as averaging or other measures of central tendency. We can use these properties to compare routines (which routine is faster/better/cheaper?). While this familiar procedure works well for properties, our current theory of organizational routines suggests that they are not just things with variable properties. Rather, they are complex generative systems that produce recognizable, repetitive, patterns of interdependent actions carried out by multiple actors (Feldman and Pentland, 2003). If we are interested in the pattern of actions – the actual performances of the routine – then familiar statistical techniques like averaging are not applicable, because the data are not scalar. Generalizing about the patterns involves induction: generalizing about a class from observing instances of that class. For example, one might observe several black crows, and on the basis of these observations, one might conclude that ‘all crows are black’. If a routine generated

50

Methods for analysing organizational routines

identical performances, the induction problem would be just that easy. Unfortunately, we know that organizational routines tend to generate a great variety of patterns. In a typical situation, we are attempting to generalize from a small set of observations to a large class that is filled with variety. Furthermore, the granularity of observations can aﬀect our impression of how much variety is present. From a distance, every performance looks the same; up close, they are all diﬀerent. One practical approach to this problem is to simply ignore the variety and describe typical performances. In process mapping, for example, we interview participants to create a flow chart (or some other kind of diagram) that summarizes their ostensive understanding of how the routine typically works. This method avoids the problem of induction because it bypasses the primary evidence – the actual performances of the routine. While flowcharts and other process maps can be useful, this seems a rather weak basis for an empirical science of routines, especially when we know that flow charts are often a poor representation of actual activities (Suchman, 1995). Properties and Patterns over Time Part of what makes organizational routines diﬀerent is their intrinsically event-based nature. To study routines up close, and understand them, we must account for the role of time. Within a single performance, time is embodied in the pattern of events. For some routines, like the invoice routine we study here, the events occur over a few hours or days. For other routines, like hiring or technology roadmapping (Howard-Grenville, 2005), events may take weeks or months. Even in a very fast routine, like a sports team driving for the goal or landing an aeroplane, timing is critical. Nevertheless, when we write about routines, these patterns are generally conceptualized as synchronic – occurring at a single moment in time. When we consider changes in the pattern over time, of course, we need a longitudinal (more macro-scale) time (Barley, 1990). Figure 4.3 shows the diﬀerent ways that time can factor into empirical studies of routines. Our familiar statistical tools are well suited for working with properties such as cycle time. Cross-sectional data take a snap-shot of a routine during a particular time window, which is suﬃcient for a ‘synchronic’ description or comparison (Barley, 1990). With longitudinal data, we can do diachronic comparisons and track how the attributes of the system change over time. When working with patterns, the empirical operations are less familiar. Typical social science methods work on scalar (or columnar) data – like questionnaire items – that can be fed into correlation-based analyses.

Using workflow data to explore the structure of an organizational routine Properties

Patterns

Cross-sectional

Synchronic comparison

Identification

Longitudinal

Diachronic comparison

Evolution

Figure 4.3

51

Approaches to identification and comparison

Network models provide a prominent example of a more sophisticated approach, and we will illustrate a network-based approach later in this chapter. Other techniques, such as classification and clustering algorithms, generally depend on computing a distance between patterns. In other words, we need to characterize the space of possible patterns and locate each observed pattern within that space. The ability to classify patterns directly would be an important aid in an empirical science of organizational routines. For example, we theorize that routines are composed of sub-routines and that these sub-routines can be re-combined in various ways to create new routines. But how can we identify particular sub-routines, given the enormous variability in a typical set of performances? The answer is not obvious, but the ability to identify and compare the structure of routines and map their change over time would be fundamental to empirical tests of our most influential theories about stability, change and evolution in organizations. Nelson and Winter (1982) hypothesize that organizations change by ‘mutation’, and by recombining routines in various ways. Genetic models and metaphors continue to be an influential approach to theorizing about routines (Becker, 2004), but empirical evidence from real organizations has been diﬃcult to obtain, and methods for analysing changes in patterns over time have been lacking. The use of workflow data may aﬀord the possibility of testing evolutionary concepts. In the following sections, we oﬀer some modest examples of how workflow data might be used to explore the structure of an invoice processing routine.

METHODOLOGY For this chapter, we collected workflow data from an invoice-management system (Compello Software) for one company over an 11-month period. The data include 2072 performances of the invoice processing routine,

Methods for analysing organizational routines

52

Optical Character Recognition

Paper Invoice

Electronic Invoice

Form Defined?

Identify and Define Form Type

Invoice Registration and Distribution

Invoice Data to Financial System

Authorization

Figure 4.4

Flow chart of the invoice processing routine

consisting of 62 756 individual actions taken from a lexicon of 31 action types. Figure 4.4 shows a flow chart for the invoice process. Invoices can enter the system on paper (which is most common) or via an electronic portal. If the invoice conforms to a known Form type, it is scanned, ‘registered’ in the system and distributed for approval. Some invoices can be immediately sent to the financial system for payment; others require multiple approvals. Once these are complete, the invoice is paid. Table 4.1 shows one performance of the invoice process routine as it appears in the workflow event log. The first column in the table indicates that the routine consists of two main phases: entry and approval. The first portion of this routine, referred to as ‘Entry’, is almost entirely automated. When ‘Person’ is ‘21’, that means that a computer is taking the action automatically, according to a rule. In the ‘approval’ phase, some actions are automatic, but others are manual. People are back in the picture. In the analysis that follows, we will sometimes refer to the routine as a whole, but at times it will be useful and interesting to compare the automated, initial entry section of the routine (‘Entry’) to the more manual, approval phase (‘Approval’). After the last approval, the invoice data is exported to a separate accounts payable system that is outside the scope of the workflow system. This is an example of a limitation to workflow event logs – their scope is limited by the design and architecture of the underlying systems.

53

Data entry Data entry Data entry Data entry Data entry Data entry Data entry Data entry Data entry Approve Data entry Approve Data entry Workflow Workflow Data entry Data entry Data entry Workflow Data entry Approve Data entry

TYPE OF ACTION CI 27475 3/31/2006 4/20/2006 33 some text 1250 2 some text 5 1250 5 33 2 2 2 Acct 16 4/5/2006 131 5 some text

DATA

1. Some columns have been removed or values changed for confidentiality

Enter document type Enter invoicno. Enter invoicedate Enter duedate Enter period Enter vendor account Enter amount Enter currency Enter text Approve Enter amount Approve Enter period Distribute to approver Distribute to approver Enter currency Enter account Enter Tax-code Notify Enter value dim. 1 (DP) Approve Enter text

Entry Entry Entry Entry Entry Entry Entry Entry Entry Entry Entry Approval Approval Approval Approval Approval Approval Approval Approval Approval Approval Approval

Note:

ACTION

Example event log from workflow system1

PHASE

Table 4.1

05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 05-Apr-06 06-Apr-06 18-Apr-06

DATE 21 21 21 21 21 21 21 21 6 21 6 19 21 6 6 21 19 19 21 19 22 44

PERSON

OCR/XML OCR/XML OCR/XML OCR/XML OCR/XML OCR/XML OCR/XML System Manual System Manual Manual OCR/XML Manual Manual System Manual System/manual System Manual Manual Manual

DATA CAPTURE

54

Methods for analysing organizational routines

In some respects, workflow data is extremely convenient, because it is computer generated. Events are unitized, categorized and time-stamped. This largely eliminates the need for the arduous and time-consuming task of coding that would be required in a typical observational or archival field study. At the same time, depending on how the workflow event log is stored, it may require some rather sophisticated computer procedures (for example, queries written in SQL) to retrieve it. Many software programs’ databases diﬀer from installation to installation, making the data diﬃcult to compare across organizations. To minimize such complications we chose a process and software which allowed a variation within a standardized framework. Once data has been retrieved, it is more or less ready for analysis. Describing the Data: Lexical Variety The most basic property of a routine is the list of events it generates. Events are fundamental because we need the events to describe and recognize the patterns. By analogy to human language, we refer to the number of diﬀerent events generated by a routine as its lexical variety or lexical size. Table 4.2 provides some measures of the lexical variety of each phase in the routine. These measures alone might suggest that ‘entry’ and ‘approval’ are basically the same – they have almost the same number of actions in their ‘lexicon’ (15 vs. 17); they have a similar average length; and a similar number of diﬀerent lexical items appear in each performance. While the number of events in each phase is similar, the events themselves are not. Figure 4.5 shows which events are found in each phases of the routine. Some events are shared between phases, yet other events are unique to a specific phase. Thus, one can easily discriminate between the phases based on the presence (or absence) of particular events. In other words, these two phases can be easily identified or recognized. In cases where there is more overlap, more elaborate classification (or pattern recognition) techniques can be used to discriminate between phases (for example, Markov and n-grams are two widely used techniques for classifying sequence data). Table 4.2

Lexical variety measures

Average length Lexicon size Average lexicon items

Entry Phase

Approval Phase

11.32 15 10.16

12.25 17 9.63

Using workflow data to explore the structure of an organizational routine

55

New scanning Forward Approved by backup Enter value dim. 4 (EM) Send back to accountant Reminder (last if more than one) Remove approver Enter Last Comment Enter value dim. 2 (PR) Enter value dim. 3 (SP) Notify Enter value dim. 1 (DP) Enter Tax-code Enter account Distribute to approver Enter document type Enter vendor account Enter invoicno. Enter invoicedate Enter duedate Enter amount Approve Enter text Enter period Enter currency 0

500 1000 1500 2000 2500 Entry Phase

Figure 4.5

Approval Phase

Frequency of action types in the entry and approval phases

Listing the Sequences Lexical variety provides a foundation, but we are ultimately more interested in the performances, which are temporally ordered patterns of events. To describe how performances of a routine vary, one could simply list all the sequences one observes. If the routine is simple enough, with fixed performances, this might be a good way to describe what the routine does. Table 4.3

Methods for analysing organizational routines

56

Table 4.3

Ten most frequent sequences for the entry phase

#

% of total

Action Sequence (119 unique, n 5 1998)

904 419 147 121 37 36 30 25 24 17

45.25 20.97 7.36 6.06 1.85 1.80 1.50 1.25 1.20 0.85

9, 12, 11, 10, 14, 21, 7, 8, 16, 2, 7 9, 12, 11, 10, 14, 21, 7, 8, 2, 7, 16 9, 12, 11, 10, 14, 21, 7, 8, 16, 2, 7, 13 9, 12, 11, 10, 14, 21, 7, 8, 2, 16, 7 9, 12, 11, 10, 21, 7, 8, 16, 2, 7, 14 9, 12, 11, 10, 14, 8, 21, 7, 16, 2, 7 9, 12, 11, 10, 14, 21, 7, 8, 2, 7, 16, 13 9, 12, 11, 14, 21, 7, 8, 16, 2, 7, 10 9, 12, 11, 10, 14, 21, 7, 8, 2, 7, 13, 16 9, 12, 11, 10, 14, 21, 7, 8, 16, 2, 7, 7

Table 4.4

Ten most frequent sequences for the approval phase

#

% of total

Action Sequence (902 unique, n 5 2297)

138 92 51 44 44 37 33 32 27 25

6.01 4.01 2.22 1.92 1.92 1.61 1.44 1.39 1.18 1.09

2, 14, 5, 8, 16, 6, 15, 20 14, 5, 5, 8, 25, 19, 2, 16, 6, 15, 17, 2 2, 14, 5, 5, 8, 25, 19, 2, 16, 6, 15, 17 2, 14, 5, 5, 8, 25, 19, 2, 6, 15, 17, 16 2, 14, 5, 5, 8, 25, 2, 16, 6, 15, 17 2, 2, 14, 5, 5, 8, 16, 6, 15, 25, 17, 19 2, 14, 5, 5, 8, 25, 2, 6, 15, 17, 16 2, 2, 14, 5, 5, 8, 6, 15, 25, 17, 19, 16 2, 14, 5, 8, 16, 6, 15, 17 2, 14, 5, 5, 8, 16, 6, 15, 25, 17, 19, 2

and Table 4.4 show the ten most frequent sequences for each phase of the routine. These sequences were produced by assigning an integer code to each event type in the lexicon (taken from the second column of Table 4.1). In the invoice routine, raw enumeration is apparently not a useful way to summarize the routine: there is too much variety. There are 119 distinct sequences in the entry phase and 902 distinct sequences in the approval phase. For the entry phase, the ten most common sequences account for 88 per cent of the observed sequences. For the approval phase, the ten most common account for just under 23 per cent. One implication of Table 4.3 is that, even for automated workflows such as the entry phase, there can be quite a lot of variability in the performances. For the less automated approval phase, variability is even higher. For this routine, collecting a small sample of performances is not likely to result in a good description. Even for the entry phase, which is almost

Using workflow data to explore the structure of an organizational routine

57

entirely automated, there are over 100 distinct sequences. In the context of the invoice processing routine, these variations may or may not be particularly meaningful. In a more complex routine, there may be more than two phases (or sub-routines), and they may be combined and recombined in many diﬀerent ways. Clearly, we need tools to identify and compare such variations and combinations.

STRUCTURE OF ROUTINES To assess the structure of the routine and possible changes over time, we need to analyse the performances. We would like to be able to compute properties of the routine that are sensitive to the patterns, so that we can perform cross-sectional and longitudinal comparisons. Better yet, we would like to be able to compare the patterns directly to support identification of routines and testing of evolutionary models. Sequential Variety To create a meaningful comparison that takes into account the pattern of events, we need to use methods that use information about the sequences. Sequential variety (Pentland, 2003) uses all of the data in a set of observed sequences (not just pairs of events) to create a single, scalar measure of the variability in the sample. It provides an index of the variability in the performances that can be used for either cross-sectional or longitudinal comparisons. Sequential variety is closely correlated to the complexity of the observed sequences, as measured by the Lempel–Ziv complexity.1 Sequential variety is computed by comparing each sequence in a sample to every other sequence (details are described in Appendix 4.1). The distance between sequences is computed using optimal string matching (Sankoﬀ and Kruskal, 1983; Abbott, 1995). This distance (also called a Levenshtein distance) can be used for subsequent clustering and pattern classification. Sequential variety is simply the average distance between sequences in a sample. Like the Lempel–Ziv complexity, this method uses entire sequences (not just pairs of actions). Figure 4.6 shows the sequential variety of the two phases of the invoice process over the 11-month period in our dataset. Figure 4.6 shows the sequential variety of each phase in the invoice routine. The numbers in Figure 4.6 are computed month by month. This is, of course, an artificial way of partitioning the data. But sequential variety depends on comparing sequences within a sample, such as a month’s worth of invoices. Figure 4.6 demonstrates that the entry phase is consistently less variable than the approval phase. This makes sense, because the entry

Methods for analysing organizational routines

58

Sequential variety over time

8 7 6 5 4 3 2 1

Entry phase

Approval phase

0 1

2

3

4

5

6

7

8

9

10

11

Month

Figure 4.6

Entry phase is less variable than the approval phase

phase is more highly automated and (as shown earlier) the sequences are more similar. Network Graphs and Measures The simplest way to summarize a large number of patterns is to break each sequence of action into pairs of actions. In this way, workflow data can be processed and displayed as a valued, directed network graph, as shown in Figure 4.5. Each node in the graph represents a type of action, while the arcs represent the sequential relationship between actions. Formally, these graphs are first-order Markov models, except that the nodes are events rather than states. Pentland (1999) calls this kind of graph an ‘action network’ because it indicates which actions follow which other actions. The action network perspective allows us to use methodologies that had previously been used for the study of social networks to ask new questions about the nature of the routine in the organization. If we graph the relationships between actions in each phase, we can visually compare how these graphs may be similar or diﬀerent. Figure 4.8 shows the network of actions found in the initial entry phase, while Figure 4.9 refers to the approval phase. The thickness of the arcs is an indication of how often a particular sequence occurs, thicker lines indicating more frequent connections between those actions. From first glance, it is easy to see diﬀerences between them. The approval phase is much more interconnected and varied than the entry phase. This is consistent with the comparison of sequential variety in Figure 4.6. More numerous (denser) connections should be correlated with higher sequential variety.

Using workflow data to explore the structure of an organizational routine

59

3 1 27 24 19 23 9

12

16 20

8 25 21

15 2

22

6

4

7

17

14

11 5 18 10

26

13 28

29

Figure 4.7

Action network for entire routine (all 31 actions) 25

21

10

29

16

11 7 2

14 9 24

12

8

13

5

Figure 4.8

Action network for entry phase

Methods for analysing organizational routines

60

31 23 8 19 15 29

5

25

6

2

14 28

3 17

18

Figure 4.9

Action network for approval phase 8

8 10

10

16

14

11

7

2

11

24 13

12

10

9

30

2 13

12

11

24

7

2 13

12

4

30 Time 1

16

14 21

7

9

4

Figure 4.10

8 16

14 21

21 24

20

16

4

9

30 Time 2

Time 3

Network of action changes over time (entry phase)

Once the routine has been represented as a network, a large number of analytical tools and techniques become available (Wasserman and Faust, 1994). In addition to the overall density, we can compute centrality of particular events, block models and structural equivalence. On a practical level, we can detect potential bottlenecks and most common paths. Dynamic Networks Because we have sequences sampled over a long period of time, we can show how the pattern of action within the routine changes over time using a software application called SoNIA (Moody, McFarland and Bender-deMoll, 2005). SoNIA creates dynamic network ‘movies’ that allow visualization of changes in a network over time. On paper, we cannot display the movies, but Figure 4.10 shows three time slices from the entry phase, based on consecutive samples of 25 invoices each. From these network graphs, the human eye can determine some common structures and gain further insight into the patterns of action over time. When the full set of invoices is animated, these patterns are even more apparent. Dynamic network graphs give an exciting moving picture that allows the human brain to use its abstraction and induction capabilities to discover shapes and regularities in the movement of nodes and arcs. Time is vital

Using workflow data to explore the structure of an organizational routine

61

to our understanding of the dynamics and flexibility found in an organizational routine, and the moving network graphs give us this.

DISCUSSION The ability to identify and compare organizational routines based on large samples of performances creates some exciting new possibilities that we have only begun to explore here. In the final section of the chapter, we oﬀer some thoughts on the limits and possibilities of this approach as we currently understand it. Data, Data Everywhere. . . The most exciting aspect of workflow data is the large and growing number of diﬀerent organizations and processes for which they may be available. Workflow systems are built into enterprise software, such as SAP and Peoplesoft. In fact, basic workflow services are built into every copy of Microsoft Windows (MS Sharepoint is a workflow manager). So it is safe to say that workflow data are ubiquitous, but there may be diﬃculties preventing their universal use in research. The first challenge, of course, is getting the data. In many operational systems, logging is turned oﬀ because it would rapidly consume too much disk space. If so, then recording workflow data would require that the organization agree to activate this functionality. Depending on how the event data are stored, it may contain data that are considered confidential or sensitive. If so, it may be diﬃcult to get an organization to invest the resources required to extract sequences of events (or other abstract representations) that do not compromise the confidentiality of the organization or its members. Organizations are understandably reluctant to share this kind of data, and protracted negotiations and non-disclosure agreements may be required. Thus, while the marginal cost of a thousand extra performances may be very low, the cost of getting any data at all may still be quite high. Unless the event log is stored in a single file or database table, considerable technical skill may be required to extract the data. The second challenge is insuring that data remain comparable between organizations. Workflow software can be configured diﬀerently by each organization using it. The utility of workflow event logs depends to a great extent on what kinds of events have been recorded. The event log records events from a particular point of view, and only to the extent that programmers (and system configuration) allow. Diﬀerent events may be monitored in the event log. Diﬀerent processing rules may be used to

62

Methods for analysing organizational routines

automate diﬀerent steps in the process. These diﬀerences are interesting and potentially researchable in their own right, but they could diminish or destroy the comparability of the data in the event logs. The third challenge in working with event log data is the novelty of the methods. We have found that we are making rather extensive use of SQL and Visual Basic to create customized data analysis procedures. Some analytical tools exist (for example, for string matching), and others are being developed. One of the most promising developments is the availability of an XML-based standard for representing workflow data and an open-source software platform for analysing these data structures (Van Dongen and van der Aalst, 2005) . This platform, called ProM, supports the use of ‘plug-ins’ for analysis. While the availability of such tools lowers the barrier to entry somewhat, analysing workflow data requires skills that are not part of a typical social science PhD program. Identifying Structure is Important While workflow data presents some challenges, the potential payoﬀ is significant. For example, it can provide a better understanding of the underlying structure of routines. For a simple case like invoice processing, detailed analysis of the workflow might be perceived as overkill. But even in our simple example, some interesting issues arose. For example, is ‘invoice processing’ one routine or two? How many routines are generating the patterns we observe and how similar are they? Based on the lexicon, the entry and approval phase seem to be very similar. They have almost the same size lexicon, and nearly the same number of steps per performance. However, the actions in each phase are diﬀerent and there is little overlap between the actions. Furthermore, the phases are very diﬀerent in sequential variety and in the actual patterns of action. Based on this analysis, ‘invoice processing’ might be better characterized as two routines. Additional sub-routines of these may be discovered with additional analysis. Combinatorics of Organizational Evolution This minor diﬀerence in description is important for a variety of reasons. First, organizations are hypothesized to adapt and evolve by combining and recombining routines. Routines are like genetic material, or perhaps building blocks, whose combinations can create meaningful diﬀerences in the overall organization. But what is the level of granularity at which these combinations can occur? Is it ‘high-level’ processes, such as ‘order-tocash’? Or is it the nominal process we see in the process map (for example,

Using workflow data to explore the structure of an organizational routine

63

‘invoice processing’)? Or is it at an even lower level, such as the ‘entry’ and ‘approval’ phases identified here? By identifying distinct routines, we begin to set some boundaries on the combinatorics of organizational adaptation at the level of routines. While workflow event logs can help identify meaningful sub-routines, they may not be very reliable as a tool for observing novel combinations and re-combinations. This is because the event log is generated by a technical artefact (a software system) that could be swapped in (or out) of the picture at any time. It would only be possible to observe re-combinations at a lower level of granularity and a shorter time-scale than the life-span of the system itself. Sequential Variety may Indicate Evolutionary Potential Such smaller-scale variations could be conceptualized as evolution within a routine as well. Feldman and Pentland (2003) suggest the possibility of variation and selective retention within routines, as exception handling and improvization create variations that may be recognized and retained as a part of the routine. Variability in routines is important because it creates opportunity for change, as new ways of doing things are tried out (accidentally or intentionally) and incorporated into the organization’s repertoire of action. To observe this evolutionary process, we need detailed data on the patterns of action in each performance, over time. Of course, workflow data oﬀer just such an opportunity. We expect that routines with higher sequential variety would have a greater possibility of displaying evidence of variation and selective retention. For example, the approval phase has much more variability than the entry phase, as shown in Figure 4.4. If the organization is able to ‘select’ desirable patterns and ‘retain’ them, they should be able to achieve improved outcomes over time. For organizations that wish to direct their evolution intelligently, the identification of patterns, causes and outcomes is necessary to selectively retain the best patterns for execution in specific circumstances. Antecedents and Consequences Workflow patterns can also be linked to outcome or performance variables. Outcome measures, such as cycle time, are often available from workflow systems, using the same data collection techniques as described earlier. By utilizing such measures we could answer questions about the consequence of sequential variety and particular workflow patterns. Potentially, one could investigate the criteria that seem to drive selection of routines (cost, quality, cycle time, and so on).

64

Methods for analysing organizational routines

It is also possible to study the antecedents of the structure of routines. Do diﬀerent environments and organizations tend to produce the same patterns, or are there systematic diﬀerences? Do diﬀerent organizations, given similar environments, produce similar patterns? Are there characteristics of the persons or team responsible for the workflow system that may predict variation in patterns of actions? In other words, are routines shaped more by the external environment or by internal features of the organization? While answers to these questions seem a long way oﬀ at the moment, the ability to analyse workflow data lets us contemplate them in ways that were never possible before.

APPENDIX 4.1

STRING MATCHING DISTANCE (ADAPTED FROM PENTLAND, 2003)

This measure is defined as the average distance between each pair of observed sequences. A standard technique for measuring the distance between two sequences that may vary in length is called optimal string matching (Sankoﬀ and Kruskal, 1983; Abbott and Hrycak, 1990; Gribskov and Devereux 1992; Sabherwal and Robey, 1993). String matching has been used extensively in molecular biology to compare protein sequences, such as DNA. Abbott (1995) provides a review of applications in the social sciences. The distance between two strings can be computed by counting up the number of operations needed to transform one string into the other. The operations include substituting one element for another, or inserting or deleting elements. Each operation has a cost, and the distance between the strings is the total cost. In this chapter, all of these costs were set equal to one, but could be adjusted to account for similarity of actions, as discussed below. The technique is called ‘optimal’ string matching because it finds the lowest cost set of operations to accomplish the transformation, thus insuring that the computed distances are unique and well-behaved (for example, they obey the triangle inequality: d(A, B) 1 d(B, C) >5 d(A, C)). Distances computed in this way are called Levenshtein distances (Sankoﬀ and Kruskal, 1983). Observations can be represented in an N 3 M array of events, where each row corresponds to one iteration of the process: e11 e Observed sequences 5 S 5 ≥ 21 ... eN1

e12 e22 ... eN2

e13 e23 ... eN3

. e24 ... eN4

. ... . ...

. e2M ¥ . eNM

(4.1)

Using workflow data to explore the structure of an organizational routine

65

where N 5 the number of observed sequences and M 5 number of events in the longest sequence. Since the length of the observed sequences may vary, this array can have a ‘ragged’ edge (signified in equation 4.1 by ‘.’). This representation includes each observation in its entirety. To estimate the variation in a set of sequences like those in Table 4.1, we can compute the distance between each sequence and every other sequence. If the sequences were all identical, then distances would all be equal to zero. If the sequences diverged from each other in a single element (for example, ‘aaa’, ‘aba’), then the distances would all be equal to one. As the diﬀerences between the sequences become more pronounced, the distances increase. Thus, a convenient and meaningful measure of variety in a set of sequences is simply the average of distances between all pairs of observations: Average distance 5

N N 1 a a d ( i, j ) n ( n 2 1 ) /2 i51 j51

(4.2)

where N equals the number of observed sequences and d(i, j) equals the Levenshtein distance between each sequence. The factor n(n 2 1)/2 is simply the number of pairs in a set of n sequences.

NOTE 1. Algorithmic complexity is a well-established approach to detecting and measuring spatiotemporal patterns in sequence data (Kaspar and Schuster, 1987). The Kolgomorov– Chaitin Complexity (often called the Kolgomorov complexity) refers to the length of the shortest self-terminating program required to produce a given sequence. While the Kolgomorov–Chaitin measure has great theoretical value, it is not readily computable (Kolgomorov, 1965). To address this problem, Lempel and Ziv (1976) devised an algorithm that provides an index of the Kolgomorov complexity. The Lempel–Ziv measure has similar properties to the Kolgomorov–Chaitin measure, and thus provides a practical measure of algorithmic complexity on empirical data (Kaspar and Schuster, 1987). For example, Butts (2001) uses the Lempel–Ziv measure to explore the eﬀects of complexity in social networks.

REFERENCES Abbott, A. (1995), ‘Sequence analysis: new methods for old ideas’, Annual Review of Sociology, 21, 93–113. Abbott, A. and A. Hrycak (1990), ‘Measuring resemblance in sequence data: an optimal matching analysis of musicians’ careers’, The American Journal of Sociology, 96(1), 144–85. Barley, S.R. (1990), ‘Images of imaging: notes on doing longitudinal field work’, Organization Science, 1(3), 220–47.

66

Methods for analysing organizational routines

Basu, A. and A. Kumar (2002), ‘Research commentary: workflow management issues in e-business’, Information Systems Research, 13(1), 1–14. Becker, J., M. zur Muehlen and M. Gille (2002), ‘Workflow application architectures: classification and characteristics of workflow-based information systems’, in L. Fisher (ed.), Workflow Handbook, Lighthouse Point, FL, USA: Future Strategies, pp. 39–50. Becker, M.C. (2004), ‘Organizational routines: a review of the literature’, Industrial and Corporate Change, 13(4), 643–78. Butts, C.T. (2001), ‘The complexity of social networks: theoretical and empirical findings’, Social Networks, 23, 31–71. Chi, M.T.H., P.J. Feltovich and R. Glaser (1981), ‘Categorization and representation of physics problems by experts and novices’, Cognitive Science, 5, 121–52. Feldman, M.S. and B.T. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48(1), 94–118. Gribskov, M. and J. Devereux (1992), Sequence Analysis Primer, New York, USA: Freeman. Haerem, T. and D. Rau (2007), ‘The influence of degree of expertise and objective task complexity on perceived task complexity and performance’, Journal of Applied Psychology, 92(5), 1320–31. Howard-Grenville, J.A. (2005), ‘The persistence of flexible organizational routines: the role of agency and organizational context’, Organization Science, 16(6), 618. Kaspar, F. and H.G. Schuster (1987), ‘Easily calculable measure for the complexity of spatiotemporal patterns’, Physical Review A, 36(2), 842–8. Kolgomorov, A.N. (1965), ‘Three approaches to the definition of the concept “Quantity of Information”’, Problems in Information Transmission, 1, 3–11. Lempel, A. and J. Ziv (1976), ‘On the complexity of finite sequences’, IEEE Transactions on Information Theory, 22. Moody, J., D. McFarland and S. Bender-deMoll (2005), ‘Dynamic network visualization’, American Journal of Sociology, 4, 1206–41. Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Harvard University Press. Pentland, B. (1995), ‘Grammatical models of organizational processes’, Organization Science, 6(5), 541–56. Pentland, B. (1999), ‘Organizations as networks of action’, in J.A.C. Baum and W. McKelvey (eds), Variations in Organization Science: Essays in Honor of Donald T. Campbell, Thousand Oaks, CA, USA: Sage. Pentland, B. (2003), ‘Sequential variety in work processes’, Organization Science, 14(5), 528–40. Pentland, B. and M. Feldman (2005), ‘Organizational routines as a unit of analysis’, Industrial and Corporate Change, 14(5), 793–815. Sabherwal, R. and D. Robey (1993), ‘An empirical taxonomy of implementation processes based on sequences of events in information system development’, Organization Science, 4(4), 548–76. Sankoﬀ, D. and J.B. Kruskal (1983), Time Warps, Strings Edits, and Macromolecules: The Theory and Practice of Sequence Comparison, Reading, MA, USA: AddisonWesley. Suchman, L. (1995), ‘Making work visible’, Communications of the ACM, 38(9), 56–64.

Using workflow data to explore the structure of an organizational routine

67

van der Aalst, W. and A.J.M.M. Weijters (2004), ‘Process mining: a research agenda’, Computers in Industry, 53(3), 231–44. van der Aalst, W., B.F. van Donger, J. Herbst, L. Maruster, G. Schimm and A.J.M.M. Weijters (2003), ‘Workflow mining: a survey of issues and approaches’, Data & Knowledge Engineering, 47(2), 237–67. Van Dongen, B.F. and W.M.P. van der Aalst (2005), ‘A meta model for process mining data’, paper presented at the CAiSE’05 Workshops. Van Dongen, B.F., D.J. Medeiros, H.M.W. Verbeek, A. Weijters and W.M.P. van der Aalst, (2005), ‘The ProM framework: a new era in process mining tool support’, paper presented at the 26th International Conference on Applications and Theory of Petri Nets (ICATPN 2005), Heidelberg. Wasserman, S. and K. Faust (1994), Social Network Analysis: Methods and Applications, Cambridge, UK: Cambridge University Press.

5.

The contribution of event-sequence analysis to the study of organizational routines Carlo Salvato

INTRODUCTION Organizational routines hold a central position in organization theory. Decades of use have shown the merits of the routine concept in developing behavioral approaches as opposed to orthodox neoclassical, stylized views of the firm. Routines have played a central role in bridging separate but complementary perspectives such as the behavioral theory of the firm (for example, Cyert and March, 1963; Winter, 1986), tacit knowledge (Polanyi, 1962, 1967), bounded rationality and limited cognition (for example, Simon, 1955), Schumpeterian perspectives on growth (for example, Teece et al., 1997). Moreover, routines have proved fruitful in laying ground for the development of new conceptual frameworks, such as evolutionary theory (Nelson and Winter, 1982), the competence perspective (for example, Winter, 2003; Dosi et al., 2000), the knowledge-based view (Kogut and Zander, 1992; Levitt and March, 1988), the dynamic capabilities perspective (Eisenhardt and Martin, 2000; Teece et al., 1997). The conceptual and empirical lens oﬀered by organizational routines is crucial in understanding how organizations accomplish their tasks in society, and how such ability evolves over time. Routines can be interpreted as the basic components of organizational behavior and as repositories of organizational capabilities (March and Simon, 1958; Cyert and March, 1963; Nelson and Winter, 1982). Hence, they are particularly relevant in understanding how such behavior changes and how capabilities are adapted to match environmental dynamism. Increasingly, routines are seen as generative, dynamic systems characterized by complex internal structures and dynamics, rather than as static, inertial objects (Pentland and Rueter, 1994; Cohen et al., 1996; Lazaric, 2000; Lazaric and Denis, 2001; Pentland and Feldman, 2005). However, a broader understanding of routines and of how organizational 68

The contribution of event-sequence analysis

69

change comes about as a result of routines evolution has been hampered by the prevalence of a ‘black-box’ approach to the study of routines and work processes in general (Pentland and Feldman, 2005: 800–801). Attention has typically been focused on the inputs and outputs of work processes, more than on process per se (Pentland, 2003a, b). While understanding process eﬃciency and eﬀectiveness is obviously central to both organizational and economic approaches, failing to pay explicit attention to the internal structure of processes used to transform inputs into outputs may result in underspecified theories of organizational change (Newell, 1990; Pentland and Feldman, 2005; Sutton and Staw, 1995). In this chapter I suggest a detailed methodological approach to unpack the structure and evolution of routines. The approach is based on the performative aspect of routines, that is, on an interpretation of routines as actual performances by specific people, at specific times, in specific places (Feldman and Pentland, 2003; Pentland and Feldman, 2005). By systematically mapping and comparing several ‘performances’ of the same routine over time – operationalized as recurring sequences of actions – the evolution of the routine can be reliably traced. Optimal matching analysis (OMA) oﬀers the algorithms which enable this complex analytical task to be performed in a relatively parsimonious way (MacIndoe and Abbot, 2004). This approach to the study of routines has already been applied in sociology to the investigation of musicians’ careers (Abbot and Hrycak, 1990) and the formation of professions (Abbott, 1990), and in organizational studies to address processes of information systems implementation (Sabherwal and Robey, 1993) and work processes (Pentland, 2003a, b). Although the approach I propose is limited to the interpretation of routines as specific actions, it lays ground for detailed interpretations of routines as broader generative systems characterized by both a performative and an ostensive part – the characterization of routines as abstract patterns that participants use to guide, account for and refer to specific performances of a routine (Pentland and Feldman, 2005: 795). Tracing the evolution of routines as interaction patterns may oﬀer an empirically grounded, systematic and reliable basis to investigate how the ostensive aspect of routine comes about and develops over time by means of specific, more or less intentional interventions by agents involved in practicing the routines, as well as by managers or external forces. Such detailed understanding of routine evolution may, in turn, strengthen the contribution of the routine construct to diﬀerent disciplinary fields and to evolutionary economics in particular (Becker, 2004, 2005; Becker et al., 2005). The chapter is structured as follows. First, I briefly review the literature focusing on the performative aspect of organizational routines, which provides solid conceptual grounding to an empirical approach based on

70

Methods for analysing organizational routines

systematic comparison of recurring sequences of action. Second, I propose a simple description of event-sequence and optimal matching analysis as a tool to unpack and study routines and their evolution. Third, I oﬀer an example of how optimal matching analysis, coupled with clustering techniques, can be used to develop an empirically-based interpretation of how productdevelopment routines have evolved in an organization over a 15-year time span. Finally, I discuss contributions and implications of event-sequence and optimal matching analysis to the study of organizational routines.

THE STUDY OF ORGANIZATIONAL ROUTINES AS EVENT SEQUENCES And that, of course, is the heart of our theoretical proposal: the behavior of firms can be explained by the routines that they employ. Knowledge of the routines is the heart of understanding behavior. Modeling the firm means modeling the routines and how they change over time. (Nelson and Winter, 1982: 128)

How can routines and their role in organizational change be investigated? The answer to this question depends on the interpretation of organizational routines accepted by the investigator. In this chapter I embrace a performative approach to routines (Feldman and Pentland, 2003), which are hence addressed as instances of organizational behavior. According to traditional evolutionary economics, routines are patterns of collective behavior, sequences of interactions among multiple organizational agents, which are followed repeatedly (for example, Winter, 1964; Nelson and Winter, 1982). The advantage of this approach is that behavior tends to be observable. It can hence be relatively easily coded and interpreted by the investigator (Becker, 2004). The alternative possibility to interpret organizational routines is to understand them as cognitive regularities or cognitive patterns (for example, Simon, 1947; March and Simon, 1958; Cyert and March, 1963), which means focusing on their ostensive aspect (Feldman and Pentland, 2003). In practical terms, this equates to investigating routines as rules for action. While several heuristics and rules of thumb (Cohen et al., 1996), standard operating procedures (Cyert and March, 1963), and programs (Simon, 1947) are made explicit by organizations through their codification, several others are held tacitly by organizational agents, and are hence rather diﬃcult to observe. It is hence essential that researchers are clear about which approach they take up in studying routines – the behavioral or the cognitive – as they imply diﬀerent epistemologies and diﬀerent analytical techniques (Lazaric, 2000). One of the advantages of understanding routines as recurrent interaction

The contribution of event-sequence analysis

71

patterns is that this allows a relatively parsimonious comparison of diﬀerent routine ‘performances’ to be carried out over time. Although a recurring stereotype sees routines as rather monolithic, fixed entities, routines are subject to relentless change pressures which shape them both from within and from their outer context. Organizational agents often follow routines without paying much attention to them. However, organizational routines are seen by an increasing number of scholars not as mindless but as ‘eﬀortful accomplishments’ (Pentland and Rueter, 1994; Feldman, 2000, 2003; Feldman and Pentland, 2003). Hence, routines are not automatically followed or reproduced – rather, organizational agents always have the choice of strictly following the rule or amending it. As Winter suggested (1985: 109): ‘Mechanistic decision making does not necessarily diminish the opportunities for genuine, deliberate choice’. This means that diﬀerent ‘performances’ of the same routine across time or space will often be characterized by more or less relevant ‘mutations’ introduced by organizational agents to adapt routine performance to local conditions and contextual contingencies. Tracing and interpreting such alterations of baseline routine will result in a detailed understanding of how routines emerge and evolve over time. Hence, systematic comparison of diﬀerent performances over time allows change to be detected in routines. Despite the rather inertial view of routines that is held by some scholars, change in routines has been a central issue since the earliest contributions, whereby a routine was defined as a ‘pattern of behavior that is followed repeatedly, but is subject to change if conditions change’ (Winter, 1964: 263). Investigating change in routines oﬀers a vantage point to understand broader organizational change. When taken as the central unit of analysis, routines interpreted as behavioral regularities oﬀer an empirically grounded approach to operationalize organizational change: ‘Because routines provide some degree of stability [. . .] they provide a contrast required to detect novelty. It is in this way that routines enable researchers to map organizational change – as incremental change of the routines themselves’ (Becker, 2004: 649). This oﬀers a compelling answer to the opening question in this section: investigating routines and their role in organizational change can be fruitfully approached by operationalizing them as patterned sequences of actions. In this chapter I propose sequence analysis and optimal matching as eﬃcient tools for analyzing routines and their evolution over time. A wide variety of work in social science concerns itself with sequences of events and phenomena. Sequence data are simply an ordered listing of items, which may be actions, numbers or anything else. Sequence analysis refers to a set of research questions about social and organizational phenomena, and a collection of analytical techniques designed to address them

Methods for analysing organizational routines

72

(Abbott, 1990; MacIndoe and Abbott, 2004). When applied to the study of organizational routines, sequence analysis allows the following questions to be addressed: ●

● ● ●

Is there a pattern, or patterns, among a set of observable sequences of actions? To what extent can observable behavior be defined as ‘routine’? If such patterns exist, how are they produced? How do routines come into being? How do patterns of event sequences – that is, routines – evolve over time? What are the consequences of routinized behavior?

Focusing on the performative aspect of routines and analyzing patterned sequences of interactions through event-sequence methods oﬀers a fruitful approach to address these questions: From such a point of view, one can then fruitfully ask why people do or do not change routines. At the same time, the proposed vantage point also provides an endogenous explanation for how the routines change, for the role of agency in performing routines, and for the fact that routines can be simple rule-following behavior at one point of time, but involve adaptive and creative behavior at another point of time. (Becker, 2004: 649)

These questions can be addressed by investigating sequence data in numerous ways. Sequence analysis methods are characterized by the fact that they treat each data sequence as a whole, rather than focusing on individual data points. The input data for sequence analysis methods are hence ordered arrays of events or phenomena. On the other hand, time series and event history analysis treat series data as generated by an underlying stochastic process, whereby individual data points are the unit of analysis (MacIndoe and Abbott, 2004). There are various types of sequence analysis methods (see Abbott, 1995 for a review). In this chapter focus is on Optimal Matching Analysis (OMA) or alignment methods. In the following section I give a short overview of OMA and of the main methodological choices it implies. I then describe an application of OMA to a sample of product-development routines.

BASICS OF OPTIMAL MATCHING ANALYSIS (OMA) Optimal matching analysis is a technique for the analysis of sequence data. As the input data for OMA are whole data sequences – rather than the

The contribution of event-sequence analysis

73

individual events comprising such sequences – OMA takes into account the full complexity of sequences. It uses an iterative minimization procedure to find the distance between every pair of sequences in a sample. Cluster analysis can then be applied to the distance matrix to ascertain if the sequences belong to distinct types drawn from a typology, that is, if there are patterns among the set of sequences. The typology, if identified, can be used as a dependent or independent variable in further analysis, to understand how patterns are produced and what their consequences are, respectively. Many interesting questions in organizational studies are about process. To address questions about organizational process, we need both longitudinal data relevant to the process of interest and technical tools to analyze those data. Longitudinal data are often available when analyzing organizational phenomena over time, and we do have a range of analytical tools at our disposal (Dale and Davis, 1994). Of the various tools for longitudinal data analysis, event history models are probably among the most useful (Blossfeld and Rohwer, 1995). They allow the researchers to test how the transition rate of moving from one state (for example, prototyping) to another (mass production) is aﬀected by other variables. Researchers can include previous history as a covariate (for example, R&D expenses). They can also include predictors which measure what is happening in other life domains (for example, market trends, suppliers), thus helping us understand the interweaving of multiple processes. However, event history models invariably consider one transition at a time. If the process in question is repeatable – as in the case when a product moves from research to prototyping, back to research, and then prototyping again – the researcher will have to model each transition separately, with the covariates suitably updated each time. This is reasonable for many purposes. But in some cases, we may want to consider the whole sequence of multiple transitions simultaneously, because we want to have a sense of how the individual steps fit together, or because we think that the sequence taken as a whole tells us something important. If this is the case, optimal matching analysis (also known as optimal alignment) is a potentially useful tool. The Background of OMA in DNA Sequencing OMA is a family of procedures that takes into account the full complexity of sequence data. The objective is to identify a typology of sequences empirically. It was developed by molecular biologists studying protein or DNA sequences. Biologists often wish to find out how diﬀerent a particular DNA sequence is from other sequences. Knowing this will help them

74

Methods for analysing organizational routines

reconstruct the evolutionary tree, or to detect relevant mutations. There are other applications of OMA in biology, other sciences and engineering (Waterman, 1995; Sankoﬀ and Kruskal, 1983). However, the origin of OMA in molecular biology – DNA sequencing in particular – bears great potential in analyzing and interpreting routines as the organizational genes, according to their original definition (Nelson and Winter, 1982: 134). Reliably and systematically tracing ‘mutations’ in otherwise apparently immutable streams of behavior may hence prove a fruitful way to address ‘the great challenge of “organizational genetics” – to understand how the continuity of routinised behavior operates to channel organizational change’ (Nelson and Winter, 1982: 135). OMA techniques were introduced into social science by Andrew Abbott, who has applied OMA to a range of sociological issues, such as the development of the welfare state and musicians’ careers (Abbott, 1995; Abbott and Hrycak, 1990). The core of OMA is a two-step procedure. First, given a set of sequences, find the distance between every pair of sequences through an iterative minimization procedure. This will give the researcher a distance matrix for all the sequences. Secondly, apply clustering or similar procedures to the distance matrix to ascertain if the sequences fall into distinct types. The typology, if found, can be used as an independent variable or dependent variable in further research. The second (clustering) step should be familiar to readers, although in the context of OMA it may run into practical problems. Hence, I will here focus on the first (computing intersequence distances) step. The OMA Algorithm In this section I will briefly explain the iterative minimization procedure through which the distances between each sequence and all other sequences within a sample are computed. The objective of this first step is to find, for each pair of sequences in the sample, the lowest ‘cost’ needed to turn one sequence into another, using three elementary operations: insertion, deletion and substitution. A simple example will illustrate this idea. Consider a hypothetical organization which originates new products by following a two-stage process of research (r) and development (d). At the end of each year, the progress of each new product is reviewed, and each product in the ‘research’ phase may be transferred to the ‘development’ phase. Here are two sequences, representing the histories of two diﬀerent products. A: r r r r d B: r d d d

The contribution of event-sequence analysis

75

B is the ‘high-flyer’: the product was ‘promoted’ to the development phase after one year, while it took A four years to do so. There are several ways to turn sequence A into sequence B. For example, we can delete the first three ‘r’s at the beginning of sequence A and then insert two ‘d’s at the end. This can be represented as follows: A: r r r r d ⵧ ⵧ B: ⵧ ⵧ ⵧ r d d d Note that ⵧ stands for an empty state – think of it as a place-holder. When it appears on the bottom line, it means that the corresponding element in sequence A is deleted. When it appears on the top line, it means that the corresponding element in sequence B is inserted into sequence A. So this matching involves three deletions and two insertions. Depending on how much a deletion/insertion ‘costs’ (see below), we can put a numerical value to this particular matching. Alternatively, we can substitute d for r for the second, third and fourth elements of sequence A, and then delete the last d. A: r d d d d B: r d d d ⵧ This matching involves three substitutions and one deletion. Again, given the costs we assign to substitution and insertion/deletion, we can attach a numerical value to this matching. What OMA does is to consider all possible ways of matching the two sequences (there are many more than the two illustrated above), and finds the cheapest way to do so. Two issues follow immediately. First, the assignment of substitution and insertion/deletion costs is a critical issue, as they aﬀect the matching outcome. How do we set these costs? Secondly, as sequences become longer, and as the number of states we wish to distinguish increases, eyeballing, as a way of finding alignment, will soon fail. For example, we may want to distinguish not just research and development, but also diﬀerent sub-stages of research and of development. We may also want to consider further upstream activities, such as technology development, or downstream activities, such as production tests, marketing tests and so on. A systematic procedure is hence needed. As far as the first issue is concerned, cost assignment is a key point where theoretical consideration comes in. Substitution costs are often set to reflect the relative distance among the states distinguished. For example, in terms of time and financial investment required, ‘research’ is closer to ‘development’ than it is to the ‘development of new technology’. Thus, it seems reasonable to set a lower substitution cost for product research

Methods for analysing organizational routines

76

b1

b2

b3

.

.

.

.

.

.

bn

0 a1 a2 a3 . . . . . . am

Figure 5.1

The matrix used in iterative minimization and part of a possible path

and product development than for technology development and product research. Researchers must rely on theoretical reasoning and external evidence to justify their cost assignment (the interested reader may refer to MacIndoe and Abbott, 2004 for more details). The second issue is about the iterative minimization procedure. The discussion here will have to be schematic. Readers can find detailed and formal discussion of the algorithm in Waterman (1995), Sankoﬀ and Kruskal (1983), MacIndoe and Abbott (2004). In general, given a finite state space S and two sequences a 5 (a1 . . . am) and b 5 (b1 . . . bn), where m is not equal to n, and the elements of a and b are drawn from S, the algorithm works by first creating a (m 1 1) 3 (n 1 1) matrix. To turn sequence a into sequence b from left to right is to move from the top left-hand cell of the matrix to the bottom right-hand cell (Figure 5.1). There are many paths one can take, but each path is built by a combination of moving down a row (which is equivalent to deleting the corresponding ‘a’ element), moving right across a column (equivalent to inserting the corresponding ‘b’ element), and moving diagonally (equivalent to making a substitution). Each cell on a particular path records the total cost incurred so far. At the beginning of the iterative process, we are at the top left-hand corner, and the cost is zero. To move down the first column is to delete successively the ‘a’ elements. Thus, the cells in the first column record the cumulative deletion costs. Likewise, the cells of the first row record the cumulative insertion costs. Given these initial cell values, we can progressively calculate the value of all cells in the matrix, by considering whether the cumulative total cost for a particular cell is minimized by moving from the cell to its left, from the cell above, or from the cell diagonal to it (see the unshaded but outlined cell

The contribution of event-sequence analysis b1

b2

b3

.

.

.

.

77 .

.

bn

0 a1 a2 a3 . . . . . . am Note:

In this example, the values of the shaded cells are known.

Figure 5.2

Progressive calculation of the minimal cost for each cell

in Figure 5.2). Having calculated the value for all cells in the matrix, we will be able to find the optimal matching for the two sequences concerned. The procedure described above produces a distance measure for every pair of sequences in the sample. Cluster analysis is then applied to the distance matrix to see if the sequences belong to a small number of distinct types. The typology generated by the cluster analysis can be used as an independent or dependent variable in further research.

AN EMPIRICAL EXAMPLE: USING OMA TO UNDERSTAND PRODUCT-DEVELOPMENT ROUTINES AND THEIR EVOLUTION I now turn to an empirical example of sequence analysis using data on product-development routines within a design firm. The aim is twofold: first, oﬀering a detailed description of all methodological steps involved in an event-sequence analysis of routines performed through optimal matching techniques; second, oﬀering insights on the potential of OMA in the investigation of organizational routines, their genesis, their evolution, and their impact on diﬀerent organizational features. Empirical Setting The study addresses the microfoundations of strategic change through the longitudinal investigation of new product development (NPD) routines

78

Methods for analysing organizational routines

within Alessi, an Italian company among the world leaders in design household objects. Alessi was founded in 1921 as a metal and lathe-works factory that manufactured small metal objects for the kitchen and the table commissioned by external clients. In the mid-1950s Alessi started to collaborate with external designers, while in the early 1970s design gradually became the driving company logic. Over the 1980s and 1990s systematic eﬀorts were directed at collaborating with external designers and architects. In those years – the focal period of this study – Alessi gradually transformed itself from a small-size firm which sporadically attracted attention of design experts with unique, though isolated objects, into a trend-setter which since then has been systematically shaping the features of its target environment. It is hence interesting to investigate how Alessi’s core routine – the product-development routine – evolved over the period which witnessed the most intense strategic adaptation of the firm. Data Sources At Alessi I have analyzed the development processes of 90 products initiated by the firm between 1988 and 2002. The tactic for data collection implied a mix of secondary and primary data, and direct observation. However, given the need to carefully trace sequences of product-development processes over 15 years, the main source of evidence was provided by access to company archives. Several hundred pages of documents for each of the 90 sampled products were available, and have been selectively transcribed, coded and analyzed. Of particular value to the study was access to ‘NPD Dossiers’, in which the firm files all documents related to each product-development process. Access to Alessi Dossiers provided extremely detailed and complete descriptions of selected NPD processes. This allowed the event sequences of investigated NPD processes to be carefully generated. The 90 projects have been selected based on a theoretical sampling logic (Eisenhardt, 1989) aimed at observing enough variability along relevant project dimensions. Given the aim of investigating the evolution of NPD routines over time, choice of the number of sampled projects in each year was critical. My choice has been to select the number of projects sampled in each year (for example, 9 projects in 1998, representing 10 per cent of the whole sample) based on the proportion of projects developed by the firm that year (that is, 206) over the total number of projects developed in the 15 years of interest (that is, 2142). This way, action sequences which took place in those years in which a higher number of projects were carried out will have a relatively higher chance of determining enduring eﬀects on the evolution of the firm’s NPD routine. The underlying behavioral

The contribution of event-sequence analysis

79

assumption is that routines develop as a result of learning from repetition (Cyert and March, 1963; Dosi et al., 2000; Nelson and Winter, 1982). The alternative strategy of drawing an even number of projects from each year (that is, six projects each year) would result in oversampling from years (for example, 1988 to 1992) in which relatively fewer products were developed. Validity and reliability of the reconstruction of the event sequences describing the 90 NPD processes rest on two qualities of the secondary data. First, each document has been methodically filed and dated by the company. Second, Alessi Dossiers thoroughly track the whole history of each NPD process, as confirmed by Alessi NPD team members. However, to ensure that Dossiers captured all relevant actions within NPD processes, early in the field observation phase I obtained from Alessi a document (developed in November 1989 by an experienced engineer) reporting a detailed sequence of all activities which are typically involved in the development of a project. When I compared this list to the sequences of activities reconstructed through NPD Dossiers, the coverage was exhaustive, since each Dossier systematically reported instances of all relevant activities listed in the document. Despite the slight chance of missing data, direct access to company documents warrants a reliability in reconstructing NPD processes which far exceeds that accomplished with alternative tactics (for example, Mintzberg et al., 1976; Sabherwal and Robey, 1993). Data Analysis Data analysis consisted of four sequential steps. In the first step I reconstructed the event sequences of the 90 product-development processes, I coded all events according to a classification scheme, and I checked the reliability of event classification. Analysis of NPD process documents – supported by selective interviews and other secondary data – allowed the event sequences for the 90 NPD processes to be carefully reconstructed. The event sequences of the 90 NPD processes yielded a total of 2897 events. This required a scheme to be devised for classifying such events into mutually exclusive and cumulatively exhaustive categories. The classification scheme (reported in Table 5.1) was developed through a number of successively refined versions, as suggested by Bakeman and Gottman (1986). A similar approach was adopted by Isabella (1990) to classify events in one organization, and by Sabherwal and Robey (1993) to classify events in 53 diﬀerent organizations. I started by developing an initial categorization scheme of 25 types of events using a subset of the 90 sequences of events. Two research

Methods for analysing organizational routines

80

Table 5.1 #

1

2

3

The classification of events within Alessi NPD processes

Event type/ Coding category Desiderata

Document which formally starts product evaluation and development Activities by Alessi Internal employees aimed at development: developing product design design Internal Activities by Alessi development: employees aimed at color definition developing new-product color(s)

4

Internal development: administrative issues

5

Internal development: production phases Internal development: packaging

6

Event description

7

Internal development: NPD process check-up

8

Internal development: critical issues raised

Activities by Alessi employees aimed at dealing with financial and administrative issues related to development of new products Activities by Alessi employees aimed at physically making a new product Activities by Alessi employees aimed at developing the packaging for a new product Activities by Alessi employees aimed at checking new-product progress along NPD phases

Activities by Alessi employees aimed at assessing and solving diﬃculties and unexpected events in newproduct development

Examples

81 Desiderata documents selectively transcribed and coded The drawings of a new product received from the designer are transformed into a detailed ‘rendering’ The ‘Color development form’ is filled in by Alessi new-product-development staﬀ for Phase 3 (out of 6 sequential phases) A cost estimate for an internally produced item is developed

The sequence and timing of the production process for a new, internally produced, item is defined Packaging for a newly developed product is selected The ‘new-product development’ form (‘Scheda SNP Prezzi’) is filled in by Alessi newproduct-development staﬀ for the ‘Capitolato’ phase Minutes of an internal meeting reporting problems and diﬃculties on a new product

The contribution of event-sequence analysis

Table 5.1 #

9

10

11

12

13

81

(continued)

Event type/ Coding category

Event description

Examples

Internal development: ad hoc approval requested Everyday contacts with designers

Activities by Alessi employees aimed at obtaining approval for non-routine activities

The Desiderata for an unusual new project is sent to the CEO for approval

Contacts between Alessi employees and external designers aimed at developing a new product

Everyday contacts with habitual suppliers: product definition Everyday contacts with habitual suppliers: administrative issues Everyday contacts with habitual suppliers: production phases

Contacts between Alessi employees and recurring suppliers*, aimed at defining product characteristics before production is started Contacts between Alessi employees and recurring suppliers*, aimed at solving administrative and financial issues

Minutes of a meeting between Alessi newproduct-development staﬀ and designers, in which the final colors for a newly developed product are discussed and decided A drawing of the product is sent to the supplier for examining productionrelated technical aspects

Contacts between Alessi employees and recurring suppliers*, aimed at producing new items (typically early phases, e.g., first trial production batch, since the study is focused on productdevelopment activities)

A cost estimate for realizing a product component is sent from the supplier to Alessi

A supplier reports that the first batch of a new product has been completed and timely delivered to Alessi A color supplier reports on the match between the developed color ‘master’ and the samples approved by the designer Alessi reports to the supplier serious faults in the last delivered batch of products

Methods for analysing organizational routines

82

Table 5.1

(continued)

#

Event type/ Coding category

Event description

Examples

14

Ad hoc modifications by external actors

Designer SG requires a preliminary study of traditional pots and pans before starting own project

15

Ad hoc modifications by Alessi personnel Contacts with unusual collaborators

Activities by external actors (i.e., excluding Alessi employees) aimed at altering established procedures Activities by Alessi employees aimed at altering established procedures

16

17

Information gathering on unusual topics

18

Others

Contacts between Alessi employees and nonrecurring collaborators*

Activities by Alessi employees aimed at gathering information about topics, issues or problems which are nonrecurring in analyzed documents Events within Alessi or in external environment that cannot be classified in the above 17 types

Alessi engineer suggests gathering plastic samples of all new colors developed for new products A study of the force required to extract a cork is required to a research lab A study of the eﬀects of garlic on colored plastic is required to a research lab Data on existing product colors used by a competitor are collected and analyzed Data on performance and price of spectrophotometers are collected by NPD staﬀ member A minority of activities not falling under any of the previous types

Note: * A list of recurring/usual suppliers was provided by the company. Coded transcripts of company documents already reported whether a supplier could be considered habitual or not. Critical assignments were solved with the help of key informants.

assistants – blind to the purpose of the research – independently classified a random set of 100 events into one of the 25 types. They agreed on the event type in only 76 cases, which was not considered satisfactory (Isabella, 1990; Sabherwal and Robey, 1993). I then repeated this procedure twice, until I developed a more parsimonious and clear classification of 18 types by creating new categories, merging categories that the independent raters perceived as similar, and by eliminating categories that coders infrequently used. Adopting this new classification scheme,

The contribution of event-sequence analysis

83

the two research assistants independently classified another random set of 100 events into the 18 categories. This time they agreed on 94 of the 100 events. The 94 per cent inter-rater agreement was compared with the chance classification accuracy – which is 4 per cent – using Cohen’s (1960) kappa. The kappa measure showed that the agreement between the two raters was significantly greater (p , 0.0001) than the agreement expected due to chance alone (Bakeman and Gottman, 1986). This procedure assured the reliability of event classification. Finally, I classified all events for the 90 NPD processes into the 18 types. To check for any intra-observer reliability decay (Bakeman and Gottman, 1997: 58–9), a research assistant coded a random sample (n 5 100) of all coded events. We agreed on the event categories in 96 per cent of the cases, and again the kappa measure was significant (p , 0.0001). In the second step I compared the 90 event sequences in terms of similarity/dissimilarity. This could be done by simple visual aids such as charts of individual event sequences, which work well when the number of sequences to be compared is small, and when they comprise few events (for example, Mintzberg et al., 1976). In my case, however, the relatively high number of examined sequences (90), and of event types (18), required the adoption of an automatic procedure to compute intersequence distances. I hence relied on optimal matching techniques, which directly measure sequence resemblance. As mentioned in the previous section, optimal matching analysis (Abbott, 1990, 1995) includes a subset of sequence analysis techniques (Bakeman and Gottman, 1997) and operates by dynamic programming, which is a class of iterative maximization techniques operating on stepwise processes. They allow distance measures to be computed among any set of sequences represented by a string of well-defined elements, drawn from a relatively small total set. The most common sequence metric establishes a ‘distance’ between sequences based on how diﬃcult it is to transform sequences into one another. The standard algorithm for alignment is the Needleman-Wunsch algorithm, which calculates alignments based on costs associated with substitution and insertion. I have analyzed the sequences of NPD codes with Andrew Abbott’s software program OPTIMIZE©.1 Once computed, these distances can be clustered, scaled, or averaged to show patterns and to reveal any common features or trajectories (Blair-Loy, 1999; Sabherwal and Robey, 1993; Pentland, 2003b). The third step in data analysis consisted in a cluster analysis of the 90 NPD sequences. Optimal matching algorithms do not directly arrange observations into sequence patterns. However, they generate interval-level measures of resemblance/distance between sequences (called Levenshtein distances; Sankoﬀ and Kruskal, 1983: 18–23), over a sequence dataset.

84

Methods for analysing organizational routines

These measures can then be used as inputs to clustering algorithms, which in turn enable the development of patterns of sequences (Abbott, 1995; Blair-Loy, 1999; Sabherwal and Robey, 1993). Cluster analysis groups the 90 NPD process sequences into five diﬀerent clusters. Each cluster represents a diﬀerent type of NPD routine, that is, a variant of Alessi productdevelopment routine. Many Alessi NPD routines seem to fall within a category which can be interpreted as the ‘standard’ way of developing a new design object at Alessi, since they closely mirror sequences dictated by internally formalized NPD procedures. All other clusters represent some kind of ‘mutation’ of the standard NPD routine (Nelson and Winter, 1982: 115–16). These mutations typically result from a combination of external and internal factors. They represent diﬀerent responses of the organization to both environmental and organizational stimuli. In other words, they oﬀer a perspective on how the Alessi product-development routine evolved in the face of internal and external dynamism, hence supporting change strategies. The final step of the analysis consisted in gradually developing an indepth understanding of the specific reasons behind each ‘mutation’ of Alessi NPD routine. Tracing each ‘mutation’ to the underlying organizational events – by either going back to the raw data, or directly asking informants through focused interviews – clarified which specific actions and actors determined the mutation within the NPD sequence. In turn, this oﬀers a detailed perspective on which actions and actors mould productinnovation routines over time, and according to what logic.

ANALYSIS AND FINDINGS The preponderance of the literature tends to describe the specific routines held by an organization as homogeneous entities. In contrast to this common view, systematic and longitudinal analysis of the microprocesses which embody a capability reveal that actual performances (Feldman and Pentland, 2003) of Alessi’s new product development (NPD) routine cluster into distinct general types. In other words, Alessi’s routines to develop new designer household objects take diﬀerent forms when ‘expressed’ into diﬀerent projects and over time. This allows a fine-grained interpretation of how Alessi’s NPD routines evolved over the period of the firm’s most intense strategic change eﬀorts. In this section I will first shortly describe the five clusters and the driving forces which determined their characterizing features. Second, I will oﬀer an interpretation of the driving logics behind routine evolution, which underpinned Alessi’s strategic change over the focal period.

The contribution of event-sequence analysis

85

Heterogeneity of NPD Routine Patterns as Revealed in Clusters I classified intersequence distances by the cluster procedure available in SAS, which allows cluster analysis to be carried out adopting distances as inputs. I explored several distance measures and clustering methods, since diﬀerent cluster algorithms yield diﬀerent solutions (Aldenderfer and Blashfield, 1984). The (no-square) Euclidean distance and Ward method yielded the most easily interpretable clusters. Adoption of the sudden jump in agglomeration coeﬃcients as a stopping rule suggested a five-cluster solution. This cluster solution was validated by an independent samples t-test. The diﬀerence between the mean of within-cluster distances and the mean of between-cluster distances was statistically significant (p ≤ 0.001). Appendix 5.1 reports the 90 NPD sequences grouped according to the five-cluster solution. Each cluster contains NPD sequences of rather diﬀerent lengths. This means that, by design, the clustering algorithm has grouped sequences based more on the nature and sequence of actions involved, than on their length. This results from the explicit choice of setting the insertion, deletion and substitution costs so that sequence length would not be the main determinant of similarity (Blair-Loy, 1999; Sabherwal and Robey, 1993). I have performed cluster interpretation by carefully analyzing cluster centroids, by going back to the raw data, which provide a rich description of each sequenced action, and by referring to secondary data – for example, product catalogues, books and publications on Alessi products and designers – or to interviews in which the focal project had been mentioned by informants. Cluster 1 (‘Recipe-book’ project sequences) Cluster 1 includes 36 NPD projects (40 per cent of the total sample). These projects represent a closer – though never perfect – match to the ‘ideal-type’ NPD sequence, than those grouped in the other four clusters. The ‘ideal type’ NPD sequence was reconstructed from a company document developed in November 1989 by the most experienced engineer in Alessi. The document reports a detailed sequence of all activities which are typically involved in the development of a project. It was developed by systematizing, over several meetings of the NPD team, Alessi’s best practices in newproduct development in previous years, and it was meant to provide formal guidance for future projects. As our data suggest, 40 per cent of the NPD processes we observed over the 1988–2002 period followed relatively closely what Alessi considers a conventional NPD sequence, as confirmed by key informants. A ‘recipe-book’ product-development routine at Alessi is typically primed,

86

Methods for analysing organizational routines

with very few exceptions, by the Desiderata (activity-type no. 1; see Table 5.1), which is the routine through which Alessi sets the basic guidelines for product-development, in line with designers’ intentions and company targets. Right after the Desiderata, ‘recipe-book’ sequences are normally characterized by contacts with the designer aimed at defining the details of the preliminary drawings on which the Desiderata is usually based (activity-type no. 10). Such contacts will recur several times throughout the sequence. Contacts with designers are followed by internal activities (no. 2) – and related external ones (no. 11) involving suppliers – aimed at developing product design in line with designers’ directions and Alessi targets. These activities allow Alessi to determine the technical details of the design through an internally-developed rendering and in close contact with suppliers of raw materials, components or – in the case of plastic products, whose production is entirely externalized – the whole product. After design details are defined, internal/external activities aimed at defining production details (no. 5 and no. 13) and administrative issues such as the cost of moulds or components (no. 4 and no. 12) follow. These preliminary phases are typically followed by a check-up of performed activities, either in the form of a revision of the Desiderata (no. 1), or a R.U.D.E. processcheck-up meeting (no. 7), or both. These check-up stages usually lead to further product-development phases, both internal (no. 2, 4, 5, and no. 3, if the product is colored) and involving external suppliers (no. 11, 12, 13). Towards the end of the process product packaging is usually decided (no. 6) and final check-up activities are carried out (no. 1, 7). Critical issues (no. 8), ad hoc approvals (no. 9), ad hoc interventions by internal (no. 15) and external (no. 14) agents, or other ‘unusual’ activities (no. 16 and 17) are only seldom present, reinforcing the interpretation of these processes as occurring in a rather ‘smooth’ and uneventful fashion. Cluster 2 (In-house mutated project sequences) Cluster 2 includes 12 NPD projects (13 per cent of the total sample). These projects show clear deviations from what can be considered a ‘standard’ NPD process at Alessi (that is, the average ‘Recipe-book’ project sequence in cluster 1, either in the nature of activities involved, or in the sequence, or both. Such variations are, in most of the instances, the result of extemporaneous behavior of agents primarily from within Alessi. ‘In-house mutated project sequences’ exhibit – relative to ‘Recipe-book projects’ in cluster 1 – a higher number of activities signaling alterations in the smooth functioning of the product development routine: critical issues raised by Alessi employees (activity type no. 8 in Table 5.1), requests for ad hoc approvals (no. 9), ad hoc modifications of ‘standard’, ‘recipe-book’ procedures introduced by Alessi personnel (no. 15) and, although less

The contribution of event-sequence analysis

87

often, contacts with unusual collaborators (no. 16) and information gathering on unusual topics (no. 17). Examples of in-house mutated project sequences include internally-derived mutations due to: ●

●

●

●

the involvement of external consultants in NPD processes due to raising product-development complications, for example, Thun’s ‘Campari’ cocktail shaker (#2, 1988), and Rossi’s ‘La Cubica’ cooking box (#8, 1991); early experiments in developing a systematic way to involve young designers new to Alessi (which will later develop in the ‘Workshop’ routine), for example, the NPD process for the ‘Hot Bertaa’ kettle (#6, 1990), or the ‘Memory containers: Creole Project’ through which objects like Cassina’s ‘Helmut’ citrus basket (#11, 1992), Boym’s ‘Tin Man’ kitchen box (#17, 1994), and Abdelkader’s ‘Gsâa’ couscous pot (#64, 1999) were developed; an experiment suggested by Alessi CEO and aimed at testing a target costing procedure in the development of new products (Castiglioni’s ‘AC04’ fruit bowl: #27, 1995); special designers’/products’ features which determined mutations in the NPD processes which yielded products such as ‘Girotondo’ (#34, 48, 58, 66, 79).

Cluster 3 (Externally mutated project sequences) Cluster 3 consists of nine NPD projects (10 per cent of the total sample). These projects have in common variations from the more standard sequences in cluster 1, which tend to result from improvisations by agents external to Alessi. These product development processes are characterized by the presence of ad hoc modifications by external actors (no. 14), and by some related contacts with unusual collaborators (no. 16) and information gathering on unusual topics (no. 17). Sometimes (for example, sequences #7, #59 and #60), these external alterations raise critical issues inside Alessi (activity type no. 8). Consequences of these alterations on the product-development process require specific internal ad hoc approval (no. 9), for example, in sequences #35, #67 and #78. Examples of externally mutated project sequences include externallyderived mutations due to: ●

suggestions by external designers to develop an innovative colorfiling system aimed at simplifying processes of new color development, for example, NPD sequences #15 (‘Firebird’), #59 (‘Dr Kleen’) and #60 (‘Rondo, Sden, Otto’), which are all plastic products having in common technical diﬃculties related to color definition;

Methods for analysing organizational routines

88 ●

●

●

external mutations resulting from special features of some designers, for example, Giovannoni and Venturini’s ‘Mix Italia’ espresso coﬀee maker (1991, #7); external mutations resulting from attempts by external designers at inverting some of the core product-development phases, due to peculiarities of the products they were asked to develop; such is the case of Giovannoni’s ‘Mami’ set of pots and saucepans (#62, 1999). The unusual procedure followed by the designer resulted in a rather altered NPD sequence, in which several internal product-development activities are carried out early in the process, while most of the designer’s activities, which are usually at the inception, follow; external mutations resulting from early attempts at developing new editions of historical Alessi projects, for example, the decorated version of Mendini’s best-selling ‘How much white’ table set in white porcelain (#35).

Cluster 4 (‘Recombinant’ project sequences) Cluster 4 comprises 20 NPD projects (22 per cent of the total sample). In contrast with the three previous clusters, these sequences encompass patterns of activities and related input flows that resulted neither from following rules which stratified before the sampled period (1988–2002), as in cluster 1, nor from improvisations performed in the course of the project, as in clusters 2 and 3. Rather, activities and patterns which seem to pool these sequences have been shaped by managerial actions intended at replicating selected improvisations, observed by managers in previous years as internal or external mutations. In other words, sequences in cluster 4 incorporate internal (cluster 2) or external (cluster 3) mutations which occurred in previous years, and which Alessi managers recognized as potentially valuable improvements to the NPD process, and eventually selected and retained. Examples of ‘recombinant’ project sequences include: ●

sequences resulting from the formalization of the Workshop procedure which emerged from early experiments at systematically involving young designers new to Alessi (falling in cluster 2). For example, projects #14, 16, 18 and 19 resulted from the ‘Family Follows Fiction (F.F.F.)’ Workshop, a more structured and refined version of the ‘Memory Containers: Creole Project’ Workshop organized some time before. Projects #26 (‘Memory Containers: Biological Project’) and #70, 71 and 72 (subsequent versions of the F.F.F. Workshop) also resulted from further refinements of the original Workshop process and structure;

The contribution of event-sequence analysis ●

●

89

sequences resulting from managerial eﬀorts at establishing a formal procedure for developing product colors, inspired by previous suggestions oﬀered by external designers (falling in cluster 3). These attempts yielded a systematic color-filing system, ‘The Color Box’, that was adopted in developing projects #28, 42, 43, 44, 52, 73 and 82; sequences resulting from a formalization of early attempts at developing new editions of historical Alessi projects (cluster 3), for example, ‘Colorbavero’ table set, #74 2001, the decorated version of Castiglioni’s ‘Bavero’ table set developed in the 1990s; ‘B 9093; 9093 GD’ kettles, #84, colored versions of the famous kettle designed by Graves in 1985.

Cluster 5 (Unconventional projects: in-house trial and error) Finally, Cluster 5 includes 13 NPD projects (15 per cent of total sample). These projects are characterized by an unconventional nature, since, with few exceptions, they are either historical reproductions (#5, 13, 25, 33, 41), or result from collaborations with unusual industrial partners or external collaborators (#32, 36, 51, 53, 56, 76, 80, 87). To sum up, the 90 sampled NPD sequences fall into a few general patterns characterized by minor within-group variation. These five patterns represent as many manifestations of Alessi NPD routine – whether a behavioral definition of routine is embraced. Forty per cent of the expressions of such routine closely mirror the NPD process sequence as formalized by top management. Nearly a quarter, the ‘in-house’ and the ‘externally mutated’, show alterations determined by improvised behavior performed by internal or external agents, respectively. Nearly a quarter, the ‘recombinant’, ensue from implementing alterations which are the result of intentional attempts at sifting out, refining, formalizing and replicating apparently promising mutations. Finally, some sequences decidedly depart from any observed NPD process regularity within the focal organization. Routines Evolving: the Interplay Between Improvisation and Intentionality Empirical evidence I have provided so far portrays an organization whose core routines emerge and develop more as a result of blind search and trial-and-error processes, than through managerial cognition. The Project Manual developed by the firm at the beginning of the focal period has apparently shaped several NPD sequences in the following years. However, it may have simply formalized practices resulting from accretion of disparate extemporaneous activities through time (Feldman, 2000, 2003). More

90

Methods for analysing organizational routines

importantly, our data point at a non-trivial number of alterations in NPD sequences, resulting from improvisational productions. This would shed a rather dim light on managerial prospects at intentionally improving organizational routines and facilitating strategic change. In spite of that, our data also show that the emergence of what we have termed ‘recombinant’ processes shows clear signs of intentional managerial intervention. Hence, I have traced – through additional secondary and primary data – seemingly intentional managerial activities through which Alessi managers selected, developed and replicated interesting ‘mutations’ which occurred in cluster 2 and 3 NPD sequences. These intentional activities always follow in time the ‘mutations’ they aim to refine and replicate (clusters 2 and 3), and precede ‘recombinant’ sequences (cluster 4), in which refined practices are later inserted. Interpretation of cluster solutions in the final stage of my analysis yielded two additional contributions. First, it allowed the interplay to be investigated between ‘random’ mutations and ‘intentional’ managerial actions in shaping organizational routines over time. As a matter of fact, NPD sequences in at least two of the five clusters appear to be resulting from ‘ad hoc’, intentional managerial actions, aimed at honing and replicating previous ‘random’ mutations. Second, it showed how the evolution of established action repertoires – whether resulting from randomness, intentionality, or a combination of the two – aﬀected the cognitive dimension of NPD routines over time, by exploring how improvised practices become gradually accepted within the organization, and how rules are shaped and become established practice.

CONCLUSION Theorists should aim to tell the truth in their theorizing, but they cannot aim to tell the whole truth. For to theorize is precisely to focus on those entities and relationships in reality that are believed to be central to the phenomena observed – and largely to ignore the rest. To advance a new theory is to propose a shift of focus, to recognize as central considerations that were previously ignored. (Nelson and Winter, 1982: 134)

This chapter was mainly aimed at illustrating the potential of eventsequence and optimal matching analysis in understanding routines seen as behavioral regularities. The empirical analysis illustrated in the previous section has explored evolutionary patterns of product-development routines in a design firm. Applying OMA to investigating routines interpreted as event sequences yielded three key findings. First, routines evolve according to distinct patterns, which imply diﬀerent combinations

The contribution of event-sequence analysis

91

of action and cognition, experimentation and logically-structured deliberation. Second, what I have termed ‘recombinant’ routines result from conscious managerial intervention aimed at replicating mutations which have already received partial positive feedback from external or internal selection forces. Hence, these managerial activities incorporate awareness of the adaptive prospects of resulting ‘recombinant’ routines. Third, ‘recombinant’ routines have higher adaptive potential as they incorporate both selective feedback of mutated routines and intentional replication activities. Hence, they constitute reliable building blocks which allow managers to constantly refocus their attention on addressing novel external and internal change stimuli. There are two main limits in this approach to studying routines. The first limit is inherent in OMA. Like any other type of analysis, OMA relies heavily on researchers’ decisions, in particular decisions about coding, and about insertion, deletion and replacement cost. Representing sequence data in a clear and parsimonious way through coding, and determining a meaningful and reliable cost scheme are central to OMA. However, results are obviously dependent on how event sequences constituting routines are represented, and on the chosen metrics. Moreover, coding routine data into event sequences and computing intersequence distances is only a first step in the analysis of sequence data. Intersequence distances yielded by OMA are mere inputs to clustering or scaling techniques aimed at discerning the existence of sequence patterns. Finally, additional interpretation is required to understand the clusters which may emerge from analysis of distance data, and to investigate the organizational eﬀects of diﬀerent routine patterns. Despite the relevant influence of the analyst’s choices, however, when carefully applied OMA may reveal patterns in routine data undiscoverable by any other method. The second limit refers to the exclusive focus on a behavioral interpretation of routines, at the expense of an interpretation of routines as cognitive regularities, currently proposed by an increasing number of scholars, either in isolation or, more often, as a complement to the behavioral interpretation. As Pentland and Feldman (2005: 793–4) suggest: ‘For many of the basic questions of organization science we need to understand the internal structure of organizational routines [. . .] When we begin to do so, we see that they are not simple, monolithic objects. They consist of both abstract understandings and specific performances’. The ostensive (abstract understandings) and performative (specific performances) aspects of organizational routines are not just relevant for descriptive purposes. Rather, they are recursive and mutually constitutive components of routines. Hence, isolating the performative aspect as OMA does, may result in a limited interpretation of routines.

92

Methods for analysing organizational routines

Focus on the performative aspect – namely, on routines interpreted as sequences of interactions – however, is not a limitation if it represents the first step in a deeper investigation of the ostensive-performative interactions. While focusing analysis on a detailed understanding of routines as event sequences has several merits in itself, it may also constitute a basis for understanding how the performative and the ostensive aspects interact. For example, it might be interesting to explore whether diversity within one aspect tends to increase diversity in the other: ‘highly contested and non-consensual ostensive aspects seem likely to produce more variation in performance [. . .] Variation in performance may also produce multiple understandings and stories’ (Pentland and Feldman, 2005: 805). A diﬀerent, but related, line of inquiry may focus on the eﬀects of divergence between the ostensive and the performative aspects: ‘A close match seems likely to indicate and predict stability and perhaps inertia. More disparate matches seem likely to indicate the existence of flexibility or change’ (Pentland and Feldman, 2005: 805). Besides the insights which can be gained by analyzing routines as event sequences, a detailed and robust understanding of the performative aspect is essential in understanding the internal structure of organizational routines as complex generative systems characterized by recursive and mutually constitutive dimensions. Hence, event-sequence and optimal matching analysis may have significant potential in advancing our understanding of routines as complex and dynamic organizational entities.

ACKNOWLEDGEMENT I am indebted to Sidney Winter for the deep and insightful conversations on the evolution of organizational routines during two extended visits to Wharton in 2003 and 2002. I am grateful to Nicolai Foss, Leif Melin and Ivo Zander for their substantial comments on previous drafts. For their useful comments and suggestions I wish to thank Andrew Abbott, Charles Baden-Fuller, Gino Cattani, Daniel Levinthal, Franco Malerba, Brian Pentland, Claus Rerup, Scott Turner, participants at the 2004 Strategic Management Society Conference, participants at ‘The Management Department Colloquia at The Wharton Business School’, and seminar participants at Jönköping International Business School, Università Cattaneo – LIUC, and Università Bocconi – CESPRI and SDA/DIR. I gratefully acknowledge financial support from Catteneo University-LIUC and from Bocconi University. This research was also supported by a grant received from the Italian Ministry of University and Research – MIUR.

The contribution of event-sequence analysis

93

NOTE 1. OPTIMIZE© is available as a convenience on Andrew Abbott’s internet site: http://home. uchicago.edu/~/abbott/om.html.

REFERENCES Abbott, A. (1990), ‘A primer on sequence methods’, Organization Science, 1, 375–92. Abbott, A. (1995), ‘Sequence analysis: new methods for old ideas’, Annual Review of Sociology, 21, 93–113. Abbott, A. and A. Hrycak (1990), ‘Measuring resemblance in sequential data: an optimal matching analysis of musicians’, American Journal of Sociology, 96, 144–85. Aldenderfer, M.S. and R.K. Blashfield (1984), Cluster Analysis, Newbury Park, CA, USA: Sage. Bakeman, R.J. and M. Gottman (1986), Observing Interaction: An Introduction to Sequential Analysis, New York, USA: Cambridge University Press. Becker, M.C. (2004), ‘Organizational routines: a review of the literature’, Industrial and Corporate Change, 13(4), 643–77. Becker, M.C. (2005), ‘A framework for applying organizational routines in empirical research: linking antecedents, characteristics and performance outcomes of recurrent interaction patterns’, Industrial and Corporate Change, 14(5), 817–46. Becker, M.C., N. Lazaric, R.R. Nelson and S.G. Winter (2005), ‘Applying organizational routines in understanding organizational change’, Industrial and Corporate Change, 14(5), 775–91. Blair-Loy, M. (1999), ‘Career patterns of executive women in finance: an optimal matching analysis’, American Journal of Sociology, 104(5), 1346–97. Blossfeld, H.-P. and G. Rohwer (1995), Techniques of Event History Modeling, Mahwah, NJ, USA: Lawrence Erlbaum. Cohen, J. (1960), ‘A coeﬃcient of agreement of nominal scales’, Educational and Psychological Measurement, 20, 37–46. Cohen, M.D., R. Burkhart, G. Dosi, M. Egidi, L. Marengo, M. Warglien and S. Winter (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5(3), 653–98. Cyert, R.M. and J.G. March (1963), A Behavioral Theory of the Firm, Oxford, UK: Oxford University Press. Dale, A. and R.B. Davis (eds) (1994), Analyzing Social and Political Change: A Casebook of Methods, London, UK: Sage. Dosi, G., R.R. Nelson and S.G. Winter (eds) (2000), The Nature and Dynamics of Organizational Capabilities, New York, USA: Oxford University Press. Eisenhardt, K.M. (1989), ‘Building theories from case study research’, Academy of Management Review, 14(4), 532–50. Eisenhardt, K.M. and J.A. Martin (2000), ‘Dynamic capabilities: what are they?’, Strategic Management Journal, 21(10–11), 1105–121. Feldman, M.S. (2000), ‘Organizational routines as a source of continuous change’, Organization Science, 11(6), 611–29.

94

Methods for analysing organizational routines

Feldman, M.S. (2003), ‘A performative perspective on stability and change in organizational routines’, Industrial and Corporate Change, 12(4), 727–52. Feldman, M.S. and B.T. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48(1), 94–118. Isabella, L.A. (1990), ‘Evolving interpretations as a change unfolds: how managers construe key organizational events’, Academy of Management Journal, 33(1), 7–41. Kogut, B. and U. Zander (1992), ‘Knowledge of the firm, combinative capabilities, and the replication of technology’, Organization Science, 3, 383–97. Lazaric, N. (2000), ‘The role of routines, rules and habits in collective learning: some epistemological and ontological considerations’, European Journal of Economic and Social Systems, 14(2), 157–71. Lazaric, N. and B. Denis (2001), ‘How and why routines change: some lessons from the articulation of knowledge with ISO 9002 implementation in the food industry’, Economies et Sociétés, 6, 585–612. Levitt, B. and J.G. March (1988), ‘Organizational learning’, Annual Review of Sociology, 14, 319–40. MacIndoe, H. and A. Abbott (2004), ‘Sequence analysis and optimal matching techniques for social science data’, in M. Hardy and A. Bryman (eds), Handbook of Data Analysis, Thousand Oaks, CA, USA: Sage, pp. 387–406. March, J.G. and H.A. Simon (1958), Organizations, New York, USA: Wiley. Mintzberg, H., D. Raisinghani and A. Théorêt (1976), ‘The structure of “unstructured” decision processes’, Administrative Science Quarterly, 21(2), 246–75. Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Belknap Press. Newell, A. (1990), Unified Theories of Cognition, Cambridge, MA, USA: Harvard University Press. Pentland, B.T. (2003a), ‘Conceptualizing and measuring variety in the execution of organizational work processes’, Management Science, 49(7), 857–70. Pentland, B.T. (2003b), ‘Sequential variety in work processes’, Organization Science, 14(5), 528–40. Pentland, B.T. and M.S. Feldman (2005), ‘Organizational routines as units of analysis’, Industrial and Corporate Change, 14(5), 793–815. Pentland, B.T. and H.H. Rueter (1994), ‘Organizational routines as grammars of action’, Administrative Science Quarterly, 39, 484–510. Polanyi, M. (1962), Personal Knowledge: Towards a Post-critical Philosophy, New York, USA: Harper Torchbooks. Polanyi, M. (1967), The Tacit Dimension, Garden City, NY, USA: Doubleday Anchor. Sabherwal, R. and D. Robey (1993), ‘An empirical taxonomy of implementation processes based on sequences of events in information system development’, Organization Science, 4(4), 548–76. Sankoﬀ, D. and J.B. Kruskal (1983), Time Warps, String Edits, and Macromolecules, Reading, MA, USA: Addison-Wesley. Simon, H.A. (1947), Administrative Behavior, New York, USA: Macmillan. Simon, H.A. (1955), ‘A behavioral model of rational choice’, Quarterly Journal of Economics, 69, 99–118. Sutton, R.I. and B.M. Staw (1995), ‘What theory is not’, Administrative Science Quarterly, 40, 371–84.

The contribution of event-sequence analysis

95

Teece, D.J., G. Pisano and A. Shuen (1997), ‘Dynamic capabilities and strategic management’, Strategic Management Journal, 18, 509–33. Waterman, M.S. (1995), Introduction to Computational Biology, London, UK: Chapman and Hall. Winter, S.G. (1964), ‘Economic “natural selection” and the theory of the firm’, Yale Economic Essays, 4, 225–72. Winter, S.G. (1985), ‘The case for “mechanistic” decision making’, in J.M. Pennings (ed.), Organizational Strategy and Change, San Francisco, CA, USA: Jossey Bass, pp. 99–113. Winter, S.G. (1986), ‘The research program of the behavioral theory of the firm: orthodox critique and evolutionary perspective’, in B. Gilad and S. Kaish (eds), Handbook of Behavioral Microeconomics, Vol. A, Greenwich, CT, USA: JAI Press. Winter, S.G. (2003), ‘Understanding dynamic capabilities’, Strategic Management Journal, 24(10), 991–95.

Methods for analysing organizational routines

96

APPENDIX 5.1

FIVE-CLUSTER SOLUTION OF NPD SEQUENCES AT ALESSI

Clusters 1 & 2/Side 1 n.

Product name

Year Description

SBU

y/n

Perf.

Designer

Wood Steel Steel Steel Steel

no yes yes no yes

n.a. F F n.a. F

Branzi Gehry Graves Brass Caramia

Steel Plastic Steel

no no yes

n.a. n.a. T

Carallo Graves King Kong

Plastic Wood Porcelain

yes yes yes

BS F F

Mendini Sottsass Sottsass

Steel

yes

T

Graves

Steel Porcelain Steel Miscellan.

yes no yes yes

T n.a. T T

Sapper Burkhardt Sottsass Starck

Plastic

yes

BS

Venturini

1996 Clothes-rack

Plastic

yes

BS

Venturini

1997 1997 1997 1997 1997 1998 1998 1998

Porcelain Steel Miscellan. Plastic Plastic Steel Plastic El. appl.

yes yes no yes no yes yes yes

BS F n.a. T n.a. T T F

Castiglioni Castiglioni Cavallaio Newson Santachiara King Kong Mendini Sapper

Plastic Steel Steel Glass Porcelain Plastic

yes yes yes yes yes yes

T F T T F F

Newson Mari D’Urbino . . . Castiglioni Castiglioni Feiz

Cluster 1: Recipe-book project sequences (n 5 36; 40%) 1 3 4 9 10

Scatolo Pito MGP Pentola Rio

1988 1989 1989 1992 1992

12 20 21

Sottopentola Bilancia Girotondo

1993 1994 1994

22 23 24

Anna G. MS0212 ES01

1994 1994 1994

29

Pellicano

1995

30 31 37 38

Bandung Vaso Sherazade Dedè

1995 1996 1996 1996

39

1996

45 46 47 49 50 54 55 57

Nonno di Antonio Appendino Antonio Bavero Grand Prix Mensola Titan Scolainsalata Girotondo Anna Light Cobàn

61 63 65 75 77 81

Sigma EM07 Augh Orseggi Tea for Two Clip-tree

1999 1999 2000 2001 2001 2001

40

Box Kettle Kettle Casserole Salt and pepper set Trivet Kitchen scale Press filter coﬀee maker Zamak corkscrew Salt grinder Oven-to-table dish Espresso coﬀee maker Automatic teapot Vase Jug Produced in aluminium Garlic-squeezer

Table set Cutlery set Shelve Soap dish Salad drainer Paper basket Cigar lighter Espresso coﬀee machine Wall clothes rack Gardening trowel Extensible trivet Glass set Tea set Magnetic paper clip holder

The contribution of event-sequence analysis

Clusters 1 & 2/Side 1

97

(continued)

n.

Product name

Year Description

SBU

y/n

Perf.

Designer

83

Ethno

Glass

yes

BS

Giovannoni

85 88

Hikuri Lupita

Porcelain Plastic

yes yes

T T

Venturini Mirri

90

Germano

2002 Kitchen box with lid 2002 Dishes set 2002 Bowl for dog food 2002 Shoehorn in PA

Plastic

yes

T

Harry & C.

Steel Steel Miscellan. Steel Steel Steel

yes yes yes yes yes yes

T F T F T T

Thun Starck Rossi Cassina Boym Castiglioni

Plastic Miscellan. Plastic Porcelain

yes yes yes yes

T T T BS

King Kong King Kong King Kong Abdelkader

Miscellan. Plastic

yes yes

F F

Giovannoni King Kong

Cluster 2: In-house mutated project sequences (n 5 12; 13%) 2 6 8 11 17 27

Shaker Campari Hot Bertaa La Cubica Helmut Tin man AC04

1988 1990 1991 1992 1994 1995

34 48 58 64

Girotondo Girotondo Girotondo Gsâa

1996 1997 1998 1999

66 79

Girotondo Girotondo

2000 2001

Cocktail shaker Kettle Cooking box Citrus basket Kitchen box Fruit bowl/ colander Kitchen timer Candles Ice cube moulds Container for couscous Soaps Cutting board

Clusters 1 & 2/Side 2 Seq. Product-development sequence length Cluster 1: Recipe-book project sequences (n 5 36; 40%) 32 32 27 29 37 22 23 32 34 30 34 32 30 26

1 10 2 2 10 2 10 11 3 5 13 4 4 12 13 13 1 7 2 3 3 11 4 3 12 3 6 6 10 6 1 7 1 10 10 2 10 11 5 5 13 2 18 4 12 1 7 10 10 10 2 2 11 4 4 6 12 5 5 10 13 6 1 7 1 10 2 10 10 2 2 11 8 5 13 14 5 4 12 1 7 2 11 4 12 5 5 13 6 1 7 10 10 2 2 10 11 10 2 5 1 13 4 2 12 1 7 2 2 10 10 11 4 4 12 4 10 6 1 7 1 10 10 2 2 11 3 2 5 10 13 4 4 12 10 10 1 7 2 2 10 2 3 3 2 10 11 3 4 12 6 3 5 13 6 1 7 1 10 2 2 11 5 5 5 13 4 12 7 2 11 4 4 12 5 6 10 4 7 1 10 2 2 13 11 13 4 12 8 13 2 3 11 3 4 3 3 4 3 12 7 7 1 10 10 2 2 11 13 10 10 5 13 4 15 12 13 1 7 2 11 4 4 13 12 5 6 6 13 6 13 1 7 13 1 10 3 2 10 11 5 13 4 12 10 10 1 7 3 2 3 10 10 10 11 3 4 12 5 10 13 6 6 6 1 7 10 10 1 10 3 2 2 11 10 10 5 4 13 4 12 10 10 1 7 2 11 3 3 4 2 12 5 13 6 1 7 13 1 10 2 11 10 2 10 10 5 13 13 4 18 4 13 12 1 7 2 11 4 12 10 13 5 13 13 6 13 6 1 7 4 12 1 10 2 11 5 13 14 13 4 12 1 7 2 10 10 11 4 4 12 5 13 13 16 10 13 13 6 6 13 1 7 10 9 10 10 2 2 13 1 10 10 6 11 5 13 4 12 2 1 7 2 13 11 4 12 6 5 6 1 10 10 7 1 10 2 2 11 5 10 3 4 4 8 12 10 7 11 4 12 2 2 13 2 10 6 7 12 12

98

Methods for analysing organizational routines

Clusters 1 & 2/Side 2

(continued)

Seq. Product-development sequence length 36 34 37 42 31 26 20 22 29 25 28 35 31 34 36 34 32 36 28 31 34 35

1 2 10 2 11 5 13 13 4 12 1 7 2 10 10 3 3 3 13 11 6 10 4 4 12 3 4 5 3 13 10 13 6 13 1 7 1 10 2 2 10 10 11 5 13 4 12 1 7 2 11 2 10 10 4 12 5 13 15 10 10 10 10 13 13 6 10 1 7 6 1 10 10 10 3 3 3 2 2 11 6 2 5 13 10 4 12 1 7 10 10 2 3 10 2 10 3 11 4 12 3 5 13 6 1 7 13 1 10 10 2 2 3 3 6 10 11 5 3 13 13 4 4 10 12 1 8 7 13 10 10 2 2 3 3 10 11 4 3 10 12 5 13 13 6 10 10 1 7 1 10 2 10 2 11 5 13 13 4 13 12 10 1 2 18 7 2 11 4 4 12 5 13 13 13 10 10 10 1 7 10 10 2 10 10 11 5 13 10 1 4 12 1 7 2 2 11 4 2 12 5 6 13 6 1 7 1 2 10 2 11 5 13 4 12 2 11 12 13 3 12 10 4 10 12 4 1 10 2 11 5 13 4 12 1 7 10 2 3 11 4 12 5 13 6 6 1 7 9 10 10 1 2 11 5 13 4 12 2 10 3 11 3 3 12 10 13 13 10 13 6 4 6 4 10 12 8 1 2 10 2 11 10 5 13 4 12 13 18 1 7 2 11 4 12 5 13 13 6 6 1 7 1 10 2 3 6 11 6 5 10 10 10 13 4 12 1 7 3 2 8 3 3 11 12 5 13 6 1 7 1 10 2 11 3 10 3 5 6 13 4 4 12 1 7 2 3 3 11 3 4 4 4 12 10 10 5 13 6 6 6 1 10 10 7 1 10 2 10 11 10 13 5 13 13 10 4 10 12 10 10 1 17 8 7 2 3 11 4 4 12 5 13 13 1 7 1 2 10 10 2 2 11 10 14 18 10 10 5 13 13 4 12 1 7 2 2 11 2 4 4 12 4 5 13 13 6 6 1 7 1 10 2 11 13 5 5 13 13 4 12 5 1 10 10 10 4 5 7 2 10 10 10 11 13 4 5 12 13 13 5 5 6 1 7 6 1 10 10 2 11 5 13 4 12 10 10 2 1 7 2 11 4 4 12 8 5 13 2 13 6 13 13 13 14 10 10 1 7 10 10 2 2 10 10 8 1 11 3 5 13 2 4 12 10 13 1 7 2 11 4 4 12 13 13 13 5 13 6 1 6 7 1 10 10 2 2 11 10 13 2 2 2 5 2 2 13 4 4 12 1 7 2 3 11 3 4 4 12 5 13 13 2 13 6 6 1 7 1 10 2 2 11 5 13 4 4 12 1 7 2 11 4 12 4 5 13 10 10 15 6 6 13 6 1 7 1 10 2 11 2 5 13 4 4 12 1 18 7 2 2 11 2 2 4 12 3 5 10 16 10 13 6 6 10 1 7 1 10 10 2 2 11 5 13 4 4 6 12 4 1 7 2 2 3 3 3 11 13 13 4 12 4 13 5 3 13 6 1 3 7 1 10 2 2 2 1 5 13 13 4 12 1 6 7 8 2 18 3 11 3 4 10 10 15 12 10 4 10 5 13 13 6 13 1 7

Cluster 2: In-house mutated project sequences (n 5 12; 13%) 27 34 42 31 30 30 37 34 25 35 30 36

16 10 2 1 8 8 8 9 10 4 12 5 1 7 8 8 11 11 10 9 9 6 13 13 13 1 7 16 10 9 9 16 9 10 10 2 8 8 1 10 10 5 5 16 15 15 3 1 10 4 12 12 11 13 13 8 8 9 13 7 7 10 2 8 8 9 10 4 12 8 17 18 7 5 9 7 5 12 2 5 8 8 8 9 17 12 7 5 7 7 8 8 16 5 17 17 12 7 6 5 5 6 10 1 8 8 8 9 17 16 15 15 10 15 10 4 5 5 1 7 8 8 2 4 5 5 4 5 6 7 5 5 4 10 8 8 9 1 10 17 8 16 16 10 15 15 10 5 5 4 1 7 8 5 2 4 5 4 5 6 10 7 5 4 10 1 8 8 15 9 9 4 2 2 2 15 15 9 2 5 5 1 2 2 4 8 4 55 5 6 7 4 10 7 1 8 7 9 17 7 7 16 3 7 10 9 9 7 10 4 5 5 1 7 8 18 3 9 9 2 4 5 5 4 5 6 7 5 5 4 10 10 1 7 9 17 7 7 16 3 7 10 15 9 9 10 4 5 5 1 7 9 9 2 4 5 5 4 5 6 7 5 5 4 10 1 7 9 17 7 10 3 7 15 10 9 9 10 4 5 5 1 7 8 9 9 2 4 1 7 1 8 8 8 9 17 16 15 15 10 15 4 12 12 5 11 11 10 4 11 1 8 8 2 4 12 13 13 4 6 7 13 4 10 7 1 8 8 15 10 10 9 9 4 2 15 15 15 9 2 1 7 2 2 5 5 4 8 4 5 6 6 10 4 7 1 7 9 19 7 7 9 16 3 7 15 10 9 9 7 10 4 5 5 1 7 18 3 9 9 2 4 5 5 4 5 6 7 5 5 4

The contribution of event-sequence analysis

99

Clusters 3, 4 & 5/Side 1 n.

Product name

Year Description

SBU

y/n

Perf.

Designer

Cluster 3: Externally-mutated project sequences (n 5 9; 10%) 7 15 35 59 60 62 67 78 86

Mix Italia Firebird How much white . . . Dr. Kleen Rondo, Sden, Otto Mami Girotondo on the beach Kalura Babyboop vase

1991 Coﬀee maker 1993 Fire-lighter 1996 Table set

Steel Plastic Porcelain

yes yes yes

F T T

King Kong Venturini Mendini

1998 Toothpick holder set 1998 Lid for toothpaste tube 1999 Pots 2000 Cotton jacquard towel 2001 Electric hot-plate 2002 Flower vase

Plastic

yes

F

Starck

Plastic

yes

T

Pirovano

Steel Textile

yes yes

BS F

Giovannoni King Kong

El. appl. Steel

yes yes

T T

Meda Arad

Plastic Plastic Plastic Steel Glass Plastic Plastic Plastic Plastic

yes yes yes yes yes yes yes yes yes

BS BS BS F T BS T F T

Giovannoni Venturini Giovannoni Giovannoni Caramia Giovannoni Venturini Mari Giovannoni

Plastic Steel Miscellan.

yes yes yes

BS T F

Giovannoni Sansoni Lassus

Plastic Steel Plastic

yes yes yes

F T T

Pirovano Pirovano Giacon

Plastic Porcelain Plastic

yes yes yes

T F T

Venturini Castiglioni Giovannoni

Steel

yes

T

Graves

Plastic

yes

T

Vos e Pezy

Cluster 4: ‘Recombinant’ project sequences (n 5 20; 22%) 14 16 18 19 26 28 42 43 44

Merdolino Gino Zucchino Molly Escar-gogò Black Josephine Mary biscuit Fred Worm EM01 Happy spices

1993 1993 1994 1994 1995 1995 1997 1997 1997

52 68 69

Alibaba Tralcio muto Strawbowls

1998 2000 2000

70 71 72

Te ò Canaglia Mr. Cold

2000 2000 2000

73 74 82

Okkio! 2000 Colorbavero 2001 Bunny & Carrot 2002

84

B 9093, 9093 GD Techno tales 1-Match

89

2002

Toilet brush Sugar sifter Kitchen scale Snail dish Biscuit box Biscuit box Jug Bread-basket Containers for spices Jug Tray Table centrepieces set Tea strainer Nail clipper Liquid soap dispenser Table brush Table set Kitchen roll holder Kettle

2002 Fire-lighter

Cluster 5: Unconventional products: ‘In-house trial and error’ (n 5 13; 15%) 5 13

Caccia Christy

1990 Cutlery set 1993 Sugar bowl

Hist. repr. Hist. repr.

yes yes

BS T

Caccia D. Dresser

Methods for analysing organizational routines

100

Clusters 3, 4 & 5/Side 1

(continued)

n.

Product name

Year Description

SBU

y/n

Perf.

Designer

25 32

90043 Sleek

Hist. repr. Plastic

yes yes

T T

Bauhaus Castiglioni

33 36 41 51 53

Firenze Patty2 Placentarius La caldissima The soundtrack

Plastic Steel Hist. repr. Plastic Plastic

yes no yes no yes

BS n.a. T n.a. F

Castiglioni Sapper Anonimo Thun Arad

56 76 80

HI FI ceramica Splugen Mangiauovo

Porcelain Steel Plastic

no yes yes

n.a. F T

Philips Castiglioni Venturini

87

Häagen Dazs

1995 Two tea infusers 1996 Mayonnaise spoon 1996 Wall clock 1996 Garbage disposal 1997 Rectangular tray 1997 Jug 1998 Self adhesive CD rack 1998 Hi Fi set 2001 Bottle opener 2001 Egg catcher-egg cup 2002 Ice cream cup

Plastic

yes

BS

Mirri

Clusters 3, 4 & 5/Side 2 Seq. Product-development sequence length Cluster 3: Externally-mutated project sequences (n 5 9; 10%) 30 1 10 10 14 14 2 10 7 2 5 4 5 5 5 14 8 7 1 7 10 10 14 4 5 5 6 14 14 6 7 36 10 1 3 2 3 11 11 13 13 18 4 14 14 14 18 1 17 14 8 14 11 11 4 12 18 6 17 11 12 6 6 5 14 14 7 16 42 9 1 10 9 9 9 10 10 10 10 9 10 10 10 10 10 10 1 18 18 5 10 18 18 6 4 14 7 7 6 5 18 4 4 18 12 6 9 18 18 18 18 33 10 1 10 2 14 2 5 5 11 11 12 4 5 8 6 6 13 14 6 7 1 7 14 4 12 13 13 13 6 10 14 14 7 34 2 1 11 10 2 14 10 11 14 14 10 2 12 4 5 14 13 8 1 7 7 10 14 4 12 12 13 13 6 13 7 6 14 14 34 1 10 10 1 14 14 14 8 2 17 10 2 14 5 4 14 8 5 5 7 1 7 10 10 2 14 4 5 5 14 14 7 6 7 32 14 1 10 9 16 11 14 14 3 9 11 2 3 10 2 16 16 5 12 4 14 8 7 1 7 4 6 14 14 14 6 7 28 10 10 1 2 14 14 9 16 3 10 3 2 4 5 14 7 1 3 10 10 14 16 7 4 6 14 7 14 28 1 10 14 14 2 10 2 5 4 5 5 14 8 7 1 7 14 10 10 5 4 5 5 6 6 7 14 14 Cluster 4: ‘Recombinant’ project sequences (n 5 20; 22%) 29 29 36 37 38 41 30 28 29

1 12 7 7 7 10 13 10 10 7 12 13 9 9 8 10 3 3 13 7 3 13 14 13 10 8 17 12 5 1 7 7 3 3 10 7 10 4 5 5 1 7 18 8 2 4 13 13 4 13 6 7 13 13 13 4 12 10 1 7 9 17 7 7 16 3 7 7 10 9 9 7 10 4 5 5 1 7 8 18 3 9 9 2 4 5 5 4 5 6 7 5 5 4 1 8 2 2 9 9 9 17 16 7 7 4 12 15 7 7 10 10 4 5 5 1 7 7 2 4 13 13 4 12 5 6 7 7 10 7 4 1 7 9 17 7 7 3 12 12 7 12 9 9 10 7 10 4 5 5 1 7 8 3 4 12 12 9 9 2 4 5 5 4 5 6 1 7 4 10 10 1 2 13 13 10 10 7 10 3 3 4 12 14 8 8 15 15 8 5 7 1 4 4 9 15 15 15 3 13 13 13 15 13 14 13 6 6 12 12 1 16 7 7 3 3 10 7 10 10 2 4 5 5 1 7 2 4 13 13 4 13 6 7 4 12 10 13 13 13 1 16 2 10 7 7 10 3 7 10 5 4 1 7 8 2 4 13 13 4 13 6 7 13 13 13 4 12 1 10 10 7 7 3 3 10 7 10 4 5 5 1 7 7 2 4 13 4 13 13 6 7 13 13 13 4 12

The contribution of event-sequence analysis

Clusters 3, 4 & 5/Side 2

101

(continued)

Seq. Product-development sequence length 32 29 32 31 31 27 29 26 32 34 29

1 10 2 17 16 7 7 3 3 10 7 10 4 5 5 1 7 8 2 4 13 13 4 13 6 7 13 13 13 4 12 10 1 7 9 17 7 9 16 7 10 9 9 7 10 4 5 5 1 7 18 9 9 2 4 5 5 4 5 6 7 2 2 1 10 17 7 7 11 3 10 7 10 5 5 4 12 1 7 2 4 13 13 4 13 6 7 13 13 13 4 12 10 1 10 2 17 16 7 7 10 7 10 4 5 5 1 7 14 8 2 4 13 13 4 13 6 13 4 12 10 7 13 13 1 10 10 16 7 7 11 10 7 10 4 5 1 7 2 4 4 3 3 13 13 13 6 7 13 13 13 4 1 12 7 10 1 10 7 10 7 7 3 3 4 5 5 1 7 2 4 13 13 4 13 6 7 13 13 13 4 12 1 16 2 7 7 3 10 7 10 4 5 5 1 7 8 2 4 13 13 4 13 6 7 13 13 13 4 12 10 1 16 7 3 7 3 10 7 10 4 5 1 7 4 13 13 4 2 13 6 7 13 7 13 4 12 1 17 2 2 16 7 7 3 3 10 7 10 4 5 5 1 7 2 4 13 13 4 13 6 7 5 13 13 5 4 12 10 1 7 9 2 17 7 16 7 3 7 10 9 9 10 4 5 5 1 7 3 9 9 2 4 5 5 4 5 6 7 5 5 4 10 1 17 10 7 7 10 7 4 5 5 3 3 1 7 8 2 4 13 13 4 13 6 7 13 13 13 4 12 10

Cluster 5: Unconventional products: ‘In-house trial and error’ (n 5 13; 15%) 39 43 34 39 38 31 25 30 40 42 31 33 39

18 1 5 3 17 17 5 16 9 13 12 7 3 16 2 16 16 1 3 17 4 14 17 17 13 11 12 18 13 3 12 7 16 3 3 6 12 5 13 1 17 17 5 11 11 16 16 16 3 5 5 16 9 18 18 2 13 3 17 4 17 12 7 14 18 3 3 14 12 13 12 6 13 7 3 5 13 16 3 3 1 16 16 16 5 5 16 9 13 17 17 5 2 17 17 4 1 17 12 12 11 13 7 14 6 7 5 13 18 14 12 13 12 16 1 17 17 10 10 16 16 17 18 18 12 3 9 16 16 5 6 11 5 17 10 5 4 18 3 12 12 16 16 5 5 4 6 12 18 12 5 5 4 18 1 17 17 5 5 16 9 16 16 3 5 2 11 13 12 12 7 14 18 3 17 4 17 17 3 3 5 13 16 14 12 13 12 6 13 7 3 1 11 8 16 17 17 11 16 8 9 2 12 12 7 16 16 1 17 4 17 17 8 14 18 14 12 12 6 7 5 16 1 16 2 17 17 2 5 4 4 6 16 11 6 5 11 6 16 12 3 2 17 13 18 16 18 10 12 10 8 10 9 12 12 12 18 10 17 18 7 10 17 16 16 16 16 16 16 16 17 18 16 16 16 16 8 18 10 16 16 1 18 3 3 2 4 4 6 7 8 9 9 10 11 17 11 1 10 10 6 6 11 11 17 12 12 13 13 13 16 16 16 17 17 18 16 8 1 9 8 16 16 5 16 5 16 16 16 16 8 16 16 16 16 18 16 16 18 18 18 16 16 18 16 18 8 18 16 16 16 16 16 16 1 3 9 17 17 10 10 12 16 5 10 5 4 18 5 6 11 5 17 5 4 5 6 3 16 12 12 12 17 17 10 1 18 2 3 3 4 4 9 9 10 11 13 10 10 13 1 6 17 17 6 7 8 11 11 11 1 17 12 12 16 16 17 1 10 3 3 18 2 6 7 8 9 4 4 9 10 11 17 10 10 18 6 6 11 12 12 13 11 11 17 1 1 13 16 16 17 17 18 13 13 16

2,897 events altogether

102

Methods for analysing organizational routines

Clusters 3, 4 & 5/Side 2

(continued)

Notes: 1. Column marked ‘n.’ reports product number (from 1 to 90). 2. Column ‘Year’ reports the year product was entered in catalogue, or decision was taken not to enter it in catalogue. 3. Column ‘y/n’ reports whether the product was entered in catalogue or not. 4. Column ‘Perf.’ reports sales performance: BS5Best seller (actual sales > 200% aspirations); F5Failure (actual sales , 50% aspirations); T5Troop (actual sales between 51% and 199% of aspirations). 5. Column ‘Length’ reports the length of each product-development process, measured as number of sequential actions in each sequence. 6. Column ‘Product development sequence’ reports the sequence of actions describing each product-development process (see Table 5.1 for interpretation of activities).

6.

The inheritance of organizational routines and the emergence of a firm genealogy in the fashion design industry Rik Wenting

1.

INTRODUCTION

Evolutionary economics starts from the proposition that heterogeneity in firm behaviour can be explained by organizational routines and their rather rigid nature to resist change unless changing conditions require it (Nelson and Winter, 1982; Hodgson and Knudsen, 2004). Routines are repetitive and collective in nature, and are to firms what habitual skills are to individuals. Organizational routines can be defined as recurrent activity patterns which are collective in nature and specific to the firm (Becker, 2004). Individual employees act in unison to perform routinized tasks that constitute the organization’s competitive edge. It is in this way that routines act as objects of selection in evolutionary models of industry dynamics (Klepper, 2002). Routines cannot be captured by codification alone and, similar to individual skills, consist of tacit and experience knowledge components (learning-by-doing). These aspects of routines render them diﬃcult to imitate by other firms (Teece et al., 1997). However, routine replication does take place between firms. In light of incentives to imitate ‘best practices’ in the industry, the rigidity and ambiguity of routines stem the replication and diﬀusion of routines. Within the organizational boundary, routines are more freely exchanged, as employees engage in day-to-day practice. Despite the proposition that routines are specific to the firm, the evolutionary literature does oﬀer mechanisms that allow for the transfer of (parts of) organizational routines between firms (Nelson and Winter, 1982), such as the setting up of new oﬃcies for a multi-locational firm, or the formation of joint ventures. A distinctive evolutionary mechanism of routine transfer is their inheritance from a parent firm to its spin-oﬀ companies (Klepper, 2002). This branch of the organizational routine

103

104

Methods for analysing organizational routines

literature is a relatively new, yet rapidly growing area of research. Spin-oﬀ companies are companies founded by a former employee of an incumbent. Spin-oﬀs overcome their liability of newness (Stinchcombe, 1965) by drawing upon their experiences within the industry. Klepper’s model of industry evolution (2002: 2007) proposes that spin-oﬀ entrepreneurs base the new firm’s ‘organizational blueprint’ on (parts of) their parent organization’s routines. As such, spin-oﬀ companies inherit their parent firm’s routines, albeit imperfectly. Empirically, one can construct the genealogy structure between firms from parent–spin-oﬀ relationships and start to analyse to what extent the success (or ‘fitness’) of parent firm and spin-oﬀ is correlated. It has become clear that spin-oﬀs indeed outperform other start-ups and furthermore that spin-oﬀs from successful parent firms outperform spin-oﬀs from less successful firms (Klepper, 2002; Dahl et al., 2003; Agarwal et al., 2004; Dahl and Reichstein, 2005; Cantner et al., 2006; Boschma and Wenting, 2007). This evidence strongly suggests that organizational routines are indeed being replicated by spin-oﬀs, albeit imperfectly. By constructing the genealogical structure of parent–spin-oﬀ relationships, we analyse whether spin-oﬀs outperform other start-ups and whether spin-oﬀs from successful parent firms outperform spin-oﬀs from less successful firms. Such an evolutionary analysis is particularly interesting in the context of creative industries, such as fashion design, music and film. Creative industries are supposedly less driven by collective routinized knowledge and more by individual creativity. To test our hypotheses, we collected a unique dataset of the fashion design industry covering 565 biographies of the world’s top fashion designers, covering the period 1858–2005. Diﬀerent from previous studies on spin-oﬀ entrepreneurship and routine transfer, we can trace routine inheritance back to all of the previous employers of the firm’s founder(s). This constitutes a complex genealogical graph from the start of the high fashion industry onwards. The transfer of routines from parent to progeny firms is hypothesized to aﬀect firm performance (Helfat and Lieberman, 2002; Klepper, 2007). Klepper (1996; 2002) proposes a model of industry evolution which builds upon this conjecture to include explanations for the aggregate phenomenon of industry life cycles. Many industries tend to follow distinct industry life cycle stages over time (Gort and Klepper, 1982; Klepper, 1997). Whereas various entrants with very diﬀerent capabilities can enter in the early stages of the life cycle, heterogeneity in organizational competences is argued to decline as the industry matures (Klepper, 2002). It is argued that competitive pressure translates into a selection mechanism which favours particular organizational routines over others. Better practices generate more profit, and allow some firms to grow vis-à-vis their competitors and gain market dominance. This process entails, over time, the exit

The emergence of a firm genealogy in the fashion design industry

105

of less competent incumbents and the rise of entry barriers as the average level of competence in the industry increases. As such, purely by selection and without exchange of learning between firms, we might expect a less heterogeneous firm population as an industry matures. However, the parent–progeny routine inheritance process is an important second mechanism which favours continued organizational heterogeneity over the industry life cycle. The proposition that parents transfer routines and part of their competitive power to spin-oﬀ firms, and that spin-oﬀ firms outperform other firms entails increased barriers to entry and a decline of firm heterogeneity over time. The agent of routine transfer in Klepper’s model is the individual employee. Since routines are collective by definition, it is important to understand the extent to which the concept of routine can be applied in general, and to the high fashion design industry in particular. Can we observe a routinization of work processes in creative industries? Indeed, to what extent is the artistic genius of head designers at fashion houses embedded in a routinized structure of production? How can individuals act as agents of routine replication, that is, to learn routines while working at a fashion house, and transplant routines albeit imperfectly from a previous employer to their own start-up firm? Routine transfer between parent and progeny firms has been found in various industries, such as accounting (Wezel et al., 2006), automobiles (Klepper, 2002; Boschma and Wenting, 2007), tyres (Klepper and Simons, 2000a), law firms (Phillips, 2002), and television receivers (Klepper and Simons, 2000b). Note that most of these studies have focused on manufacturing industries, and rarely on creative industries (Wenting, 2008). This may be due to data limitations, but we argue that creative industries are particularly interesting when testing the spin-oﬀ mechanism and the usefulness of the routine concept. Testing an industrial evolution theory based on the diﬀusion of routinized practices through spin-oﬀs might be especially interesting for a creative industry such as fashion design, specifically since one is hard pressed to reconcile routines’ inertia with a production process that revolves around creative and artistic genius. Indeed, routinization and creativity have been characterized as an organizational duality (Ohly et al., 2006). Understanding the relationship between routinized and creative practices is of particular importance to firms that rely on the creative and proactive behaviour of their employees. We argue that this is especially the case for cultural industries, and address these issues in our analysis of the fashion design industry. In this chapter we aim to analyse the genealogy of the fashion design industry throughout its history and examine the role of spin-oﬀ creation in the diﬀusion of routines between fashion houses. In particular we set out

106

Methods for analysing organizational routines

to uncover the role of key individuals who learn organizational routines at parent firms, only to transfer their competence to a new entrepreneurial eﬀort of their own. The remainder of this chapter is structured as follows. In the next section we start with a short history of the fashion design industry and its industrial dynamics. In addition, we discuss the life courses of some of the most influential designers and their role in the founding years of the industry. We discuss routinization and replication of organizational routines in the fashion design process. In section 3 we present our theoretical framework and propose several hypotheses on routine inheritance and spin-oﬀ formation. Our data and methodology are presented in section 4 and the results are presented in section 5. We end with some overall conclusions and discussion.

2.

THE HISTORY OF THE HIGH FASHION DESIGN INDUSTRY AND THE IMPORTANCE OF GENEALOGY

The fashion design industry started with the formation of the first couture house in 1858 by Charles Frederick Worth in Paris (De Marly, 1980). Selling clothes has long been part of the French economic system. The production of clothing for women throughout France was estimated to be worth £4 million in 1864, of which 40 per cent was made in Paris (De Marly, 1980). However, only one sixth of this amount was for export, and the sort of garments that were principally exported were several types of wraps, and embroidered accessories. That situation changed with the emergence of Worth’s new industry: haute couture. In 1914 Worth stated that Paris earned $50 million dollars a year for couture dresses alone. By 1927, the golden age of haute couture, the industry employed 250 000 employees in 2000 workrooms, salons and shops, and the value of its exports was equal to one-tenth of total French world trade. Today, the industry remains dominated by a handful of Parisian firms, but has branched out to other major fashion capitals since the mid-1960s, namely London, New York and Milan. Haute couture or high fashion diﬀerentiates itself from the clothing and textile industry through the incorporation of artistic design and technical excellence. Based on initial demand from the French imperial court, Worth’s couture created its own market by designing clothing for Europe’s royalty. Indeed, Worth’s dresses for empress Eugenie in 1860 opened up a new way of approaching the production of high fashion. Worth also made use of models to show his creations to his clientele while they visited his shop. In fact, at set times each year he showed his designs at fashion shows. No longer did the client dictate design, which was the custom up

The emergence of a firm genealogy in the fashion design industry

107

until the mid-nineteenth century. All creative freedom came to rest in the hands of the couturier, who would sign his or her work to make each dress an exclusive item which would build upon the designer’s reputation. As head designer and founder, Worth directed a team of assistant designers and other personnel responsible for cutting, sewing and the purchasing of luxurious fabrics. Indeed, the high fashion industry became a luxurious industry that provided products for social distinction and signalling which favoured both the couturier and the client. The way a couture house is set up, and its day-to-day practices, are still based on these standards set out by the industry’s founder. The share of Paris-based firms started to fall rapidly from around 70 per cent up until 1923, to 36 per cent in 1941 (O’Hara Callan, 1998; Watson, 2004). Currently fewer than 25 per cent of the famous fashion houses in the world are located in Paris. At the same time, London, New York and Milan increased their share of designers, with New York currently outnumbering Paris. The marked shift in spatial concentration of the industry is closely related to the success of prêt-à-porter or ready-to-wear in the 1950s and 1960s as a new market segment (Waddell, 2004). Ready-to-wear is a simplified form of high fashion and demands less artistic skill from the designer. Initially French haute couturiers were not allowed to practise ready-towear according to the guidelines of the Syndicate Chamber of Parisian Couture. This association, founded in 1868, regulated the profession of the French haute couturiers. As a consequence French couturiers entered the ready-to-wear market much later than designers outside Paris. And, with prêt-à-porter serving a rapidly increasing consumer group outside France, the fashion clusters in London, New York and Milan could grow much faster than the cluster in Paris, leading to a spatial de-concentration of the industry (see Figure 6.1). From its onset the high fashion industry has been serving the international market. In the last decades, the high fashion design industry has grown into a truly global industry, where fashion design houses such as Versace and Dior turn over billions of dollars, and we have seen the rise of new fashion clusters such as Hong Kong, Shanghai, Mumbai, and Los Angeles. Spin-oﬀs have only been important for fuelling the industrial population since the early 1900s (see Figure 6.2). The industry’s earliest entrants were either diversifying from other industries or had no related experience at all. The industry’s founder, Worth, was an experienced entrant, who had been working for London and Parisian textile firms. Maison Worth showed many signs of its founder’s having begun his career in a textile shop, for many fabrics were placed on display throughout the salons (De Marly, 1980). Even though both experienced and inexperienced entrepreneurs

Methods for analysing organizational routines

5

30 15 0 1940

1938

1928

1918

1908

1898

1888

1878

1868

0

45

Time (years)

Source:

2000

10

60

1990

15

75

1980

20

90

1970

25

105

1960

30

1950

Number of fashion houses

35

1858

Number of fashion houses

108

Time (years)

Paris

London

Milan

New York

LA/Hollywood

Elsewhere

Own elaboration of O’Hara Callan (1998) and Watson (2004).

Figure 6.1

The geographical structure of the global fashion design industry, 1858–2005

100

Percentage

80 Spin-off firms Experienced firms Inexperienced firms

60 40

2005

1984

1963

1942

1921

1900

1879

0

1858

20

Time (years) Source:

Own elaboration of O’Hara Callan (1998) and Watson (2004).

Figure 6.2

Composition of the fashion design firm population by pre-entry background

were able to enter the new market, experienced firms dominated the industry. Worth only faced serious competition from 1871 onwards when experienced entrant Jacques Doucet set up a salon in Paris. For a long time Doucet’s name was the only one equalled with Worth (Chapon, 1996).

The emergence of a firm genealogy in the fashion design industry

109

Worth, Doucet and many other early entrants started to generate spinoﬀs in considerable numbers after the 1920s. The share of spin-oﬀs in the number of fashion houses had surpassed the share of inexperienced firms in 1927, and that of experienced firms in 1982. Today, the majority of topcouture houses are spin-oﬀ entrants. The reason for the rise of spin-oﬀ entrants is two-fold. First of all, as the number of incumbents grows over time, logically, the number of potential entrants through spin-oﬀ does so as well. Thus, the chance of a spin-oﬀ entrant increases with time. Second, as competition filters out less successful incumbents, the share of successful spin-oﬀs also increases. Of course, this latter line of reasoning rests on the assumption that fitter incumbents either produce fitter spin-oﬀs, or that they generate spin-oﬀs more frequently (Klepper, 2002). Our descriptive data seems to support this assumption. Indeed, many descendants can trace their parentage back to a handful of successful early entrants (see Figure 6.3). One can observe that most firms start as spin-oﬀs and that most spin-oﬀs have long family traditions, some of which go back to the first founding father being the Paris-based company of Worth. Interestingly, early entrants in New York and in Milan were start-ups (Carnegie and Prada, respectively) who acted as founding fathers of many local, succesful spin-oﬀs. Figure 6.4 depicts a part of Figure 6.3, namely the descendants of the Parisian house of Piguet. Piguet generated five spin-oﬀs (Castillo, Balmain, De Givenchy, Galanos and Dior), who generated a fair number of spinoﬀs themselves, and so on. Amongst them are many famous designers such as Bohan, Gaultier, Saint Laurent and De la Renta. Note that aside from the latter, most designers are based in Paris, just like the majority of their ancestors. This signals the localized nature of the spin-oﬀ process. In general, spin-oﬀs seem to prefer to start their firm near to their parent organization (Klepper, 2002). Since Piguet trained with Poiret, who was an assistant designer to Worth and Doucet, the ‘family’ lineage can be retraced to the industry’s earliest formation. Similar to the successful Parisian designers, early entrants such as Carnegie (1909) in New York, and Prada (1913) in Milan trained many designers who would start their own firms at a later stage within the same cluster. For example, Banton (in 1924) and McCardell (in 1940) are spinoﬀs of Carnegie (with a total of seven spin-oﬀs), while Audibet (in 1984) and Alaia (in 1980) have worked for Prada (with a total of five spin-oﬀs) before starting their own company. Aside from these early entrants, the new fashion capitals emerging with the rise of ready-to-wear in the 1950s and 1960s were also fuelled by new start-ups. In Milan, serial entrepreneur Arnoldo Girombelli was especially noteworthy in fuelling the growth of the local fashion cluster. He founded several Italian ready-to-wear companies:

Methods for analysing organizational routines

110 a. b. c. d.

Worth Doucet Redfern Lucile

e. f. g. h.

Raudnitz Poiret Cheruit Balenciaga

i. j. k. l.

London

Paris

Carnegie Prada Lelong Schiaparelli

m. Schuberth n. Piguet o. Fath p. Dior

New York

Milan

1858 a

1861 b

1871

100 50 Number of descendants 0

c d

1887

e g

1900

f i

Year of entry

Elsewhere

spinoff firm Non-spinoff firm Spinoff relation

1913

h

j

k

1926

l o

1939

n

m

p

1952 1965 1978 1991 2004

7902

Source:

Own elaboration of O’Hara Callan (1998) and Watson (2004).

Figure 6.3

The genealogy of the fashion design industry, 1858–2005

Genny in 1961 and Byblos in 1973, which would come to generate 18 spinoﬀs themselves. In New York, Donna Karan (started in 1984) was herself a former employee of the New York-based fashion houses of Anne Klein and Dell’Ollio, and has generated eight spin-oﬀs. Six of these are located in New York. In short, a few successful early entrants generated many other successful couture houses, who generated many spin-oﬀs themselves, and so on, culminating in the structure of the industry we see today. Furthermore, most of the couture houses active today can trace their descent back to only a handful of early entrants in nineteenth-century Paris. It remains to be seen, however, to what extent ‘family traditions’ are indeed passed on in the creative industry of fashion design. Indeed, first of all, we need to reconcile the industry’s obsession with remaining creatively relevant (Banks, 2000) with the rigidity of organizational routines.

The emergence of a firm genealogy in the fashion design industry

111

Gaultier Castillo Cardin De la Renta Luca Montano

Balmain

Piguet

De Givenchy

Bohan

Spook

Dior

Saint-Laurent Amsler Galanos

Location:

Scherrer

Paris London New York Milan Elsewhere

Source:

Own elaboration of O’Hara Callan (1998) and Watson (2004).

Figure 6.4

Spin-oﬀ descendants of Piguet in the period 1933–2005 and their location

3.

THE SPINOFF MECHANISM OF ROUTINE REPLICATION AND CREATIVITY

3.1

The Dualism of Organizational Routines and Individual Creativity

The concept of routines underlies the evolutionary theory of the firm, as it posits firm behaviour as heterogeneous and resistant to change (Nelson and Winter, 1982). Indeed, ‘the core concern of evolutionary theory is with the

112

Methods for analysing organizational routines

dynamic process by which firm behaviour patterns and market outcomes are jointly determined over time’ (Nelson and Winter, 1982: 18). Following Becker (2004), we define organizational routines as recurrent patterns of behaviour, to the extent that they are communal and processual in nature. Note that this implies that routines of firms are dissimilar to habits of individuals, in a way just as firm capability diﬀers from individual skill (Becker, 2004; Pentland and Feldman, 2005). Since individual action is embedded in the repetitive interplay of multiple actors, routinization has been argued to harm artistic creativity and technological innovation alike (Ford and Gioia, 2000). This duality between routinized behaviour and the production and implementation of novel ideas stems from the rationale that standardization of patterns of actions narrows the range in which actors can search. As stated by Scott and Bruce (1994, p. 587): ‘The problem solver working within established methods or procedures is likely to generate conventional solutions.’ Zellmer-Bruhn (1999) argues similarly that interruptions strengthen search for external routines, while frequent and regular repetition strengthen the existing routine. This is particularly the case when time pressure is involved, as it leads to reliance on routine responses (Zellmer-Bruhn, 1999). Interestingly, even though the argument for this rationale might seem intuitive, empirical evidence is, as of yet, scarce (Ohly et al., 2006). We want to stress here, however, that even though routines may adapt to circumstances over time, the underlying search process is subject to rigid cognitive patterns itself. This aspect of search routines can be found in Nelson and Winter (1982: 17), where they state that: ‘at least from time to time some people within the firm may engage in scrutiny of what the firm is doing and why it is doing it, with the thought of revision or even radical change’. We propose that these processes, like other ones, are ‘ruleguided’. Thus, ‘improvisation’, often the basis for artistic creativity, is set against a background of rules and expectations (Weick, 1998). Although the particular courses of action are always to some extent novel, they are part of a specific set of possible actions within cognitive boundaries. Even though such a characterization of routinization and creativity can be useful in deciphering organizational performance in general, it is particularly so for analysing organizations engaged in aesthetic value production. In other words, novel action may be unique in its own right, but only to the extent that it is the product of rigid and systemic search processes. Fashion houses in the high fashion design industry function as the nexus of luxury and brand image, where commerce meets aesthetics. Production of couture involves much more than creative vision alone. A design house is a relational nexus with a host of famous clientele, suppliers of luxurious fabrics, the international press, various related industries such as photography and modelling, and retail outlets. These tasks are traditionally heavily

The emergence of a firm genealogy in the fashion design industry

Table 6.1

113

Fashion cycles and the designer’s calendar of catwalk shows

Location

Summer collection show

Winter collection show

New York London Milan Paris

Early September Mid-September End of September Early October

Early April End of February Early March Mid-March

influenced, if not handled singlehandedly, by the head designer, commonly the founder of the house. Assistant designers are allowed a fair share of responsibility, which oﬀers enough opportunity to remain creative, but also to learn the tricks of the trade. In the fashion design industry in particular, tight seasonal time schedules further strengthen the need to standardize a cyclical creative process. Indeed, couturiers need to present new collections each season (Table 6.1). Frequent and regular repetition strengthens the existing routine. As above, this is particularly the case when time pressure is involved, as it leads to reliance on routine responses (Zellmer-Bruhn, 1999). Aside from this, repetitive behaviour patterns are strengthened by the project-based nature of the design process (Vinodrai, 2006). In short, creative processes can, to some extent, be routinized. On first glance, the fast-moving and creative industry of high fashion seems the last place to find repetitive and somewhat rigid, collective behaviour which might be denoted as organizational routine. However, exactly because of the complexity of the fashion design process, the various types of external relations, and the tight time schedule, make for a six-month schedule that embeds all necessary steps towards the next catwalk show in an organizational routine. The design routine, then, is creatively adapted to new fads and fashion and the artistic vision of the head designer, without forsaking any of the traditions laid down during the foundation of the house’s image and clientele. The problem of exclusivity and brand recognition bind the scope of adjustments to the organizational routine which dominates the design process in every new fashion cycle. Of course, this form of routinization is not exclusive to fashion design. Indeed, it can be found in many other cultural industries, and in activities such as consultancy and policymaking as well. Firms active in such industries routinely create incremental changes in their products, usually at set intervals. 3.2

Spin-oﬀ Mechanism of Inheritance

Within evolutionary studies of industrial formation, spin-oﬀ dynamics has taken a prominent role (Buenstorf, 2006). This is primarily because

114

Methods for analysing organizational routines

of the reasoning that spin-oﬀs replicate (part of) their parent’s organizational routines within their own organizational framework. It rests on the assumption that new routines are hard to implement within established organizational formats, and are easier within new organizations. Pre-entry experiences of the founder(s) become imprinted within the new organizational routines or ‘blueprint’ (Helfat and Lieberman, 2002). Such an organizational blueprint is made by the organization’s founder(s). The importance of a firm’s founder for shaping its organizational structure is especially true for fashion houses. The way a couture house is set up, its departments, its personnel and its day-to-day activities are entirely based on the Parisian model (Waddell, 2004). It dictates that the head designer or couturier is the figurehead for the house, and is usually the founder him or herself. His or her name is vital to the house’s image and success, as has been the case from Charles Worth in the 1860s to Christian Lacroix today (De Marly, 1980). ‘The designer couturier sets the mood of the house, designs the collections, oversees the toils, the fittings, chooses the models, arranges the shows, talks to the press, negotiates with backers and takes full responsibility for the house, its reputation and its success’ (Waddell, 2004: 3). The designer’s reputation, label and social ties with peers and media, form the most important components for a fashion house’s competitive power. Thus, fashion houses are set up around a single individual – the founder or his/her successor. When the designer founder retires, he or she opens up a vital position within the fashion house. Most design houses can last several years. Only a few houses, however, outlast their founder. Indeed, finding a suitable replacement is diﬃcult. The clientele are loyal to a house’s style and reputation, and any new head designer has to conform with their expectations of continuity. An example is Donatella Versace’s emergence on the fashion stage, filling her brother’s shoes after his death in 1997. Even though Donatella was embraced by peers and press, the house of Versace experienced sagging sales for two years (Bawa, 2007). Most couturiers have assistant designers but their name is rarely known until they, in turn, set up their own firm. Of course, by doing so, new entrants only reinforce the established organizational structure. By imprinting the organization with their pre-entry experiences, the founder shapes the behaviour of their organization (Klepper, 2002). Notice that in our analysis routines are treated as a black box. We observe what pre-entry experiences dictate the new organizational routine and the performance thereof, and have no idea of the inner workings of the work practice itself. Since we assume that routines behave like an organizational ‘blueprint’, the ‘black-box’ approach is appropriate (Pentland and Feldman, 2005). Indeed, the usage of routine metaphors such as ‘blueprint’ and ‘DNA’, is to reflect its rigidity in shaping outcomes (Nelson

The emergence of a firm genealogy in the fashion design industry

115

and Winter, 1982). We expect that new entrepreneurs take their valuable pre-entry experiences at incumbents with them, which inherently gives them a competitive edge compared to inexperienced entrants (Helfat and Lieberman, 2002). In a similar vein, entrepreneurs with pre-entry experiences or diversifying firms that originate from related industries are also considered to outperform inexperienced entrants. We have defined fashion manufacturing, textile manufacturing and fashion or textile trade as related sectors to fashion design. Thus, the following hypotheses are formulated: Hypothesis 1: Firms with an ‘experienced entrant’ background have lower hazards to exit compared to other entrants. Hypothesis 2: Firms with a ‘spin-oﬀ entrant’ background have lower hazards to exit compared to other entrants. Furthermore, we expect that firm survival is aﬀected by the amount and type of pre-entry experience (that is genealogy) that the spin-oﬀ entrepreneurs have build up in the past. Note that in the case of fashion design, where most companies are small and rely on a single brand, the survival of a firm is a good indicator of its success. Spin-oﬀs may not only outperform other entrants because of their more related experience with the industry. Pre-entry experience in the industry does not necessarily capture the eﬀect of routine transference between parent and progeny firms. In a heterogeneous set of firms, where firm capabilities diﬀer based on their organizational routines, better-performing parents might function as more fertile learning environments for aspirant entrepreneurs. In other words, in light of routine inheritance through spin-oﬀ formation, better-performing parents should generate more and more successful spin-oﬀs (Burton et al., 2002; Klepper, 2002). By counting the number of years a parent firm had been active in the industry at the time of the spin-oﬀ’s founding year, we are able to approximate the quality of the pre-entry experience of the spin-oﬀ entrepreneur. As such, we distinguish between spin-oﬀs from more successful firms in terms of survival, from those from less successful firms. The hypothesis follows: Hypothesis 3: Spin-oﬀs of successful parents have lower hazards to exit, compared to other entrants. In addition, we might examine whether the eﬀect of pre-entry experience changes as the firm ages. The Cox hazard model assumes that the eﬀects of covariates on survival are proportional, or constant, over time. However, a firm’s initial routines, as set up by the founder, might be changed slightly by

116

Methods for analysing organizational routines

future employees and head designers in light of changes in (external) conditions (Nelson and Winter, 1982). As such, we may expect the eﬀect of routine inheritance through spin-oﬀ formation to change over time as the firm ages. In other words, if the founder has indeed laid down the blueprint of the organizational routine, what is the duration of this eﬀect? In Klepper’s (2002) model of industry evolution, firm capabilities inherited through pre-entry experiences do not change as the firm ages. The model does allow for scale appropriability, that is, an increase in the amount of resources over time. The eﬀectiveness to deal with these resources, however, remains dependent on the initial conditions set out by the founder. To test the proposition that the pre-entry experience of the founder has such a lasting eﬀect on firm survival, we incorporate a time-dependent covariate in the model. This allows us to observe the change in the relationship between inherited routines and firm survival as the firm ages. We take Klepper’s model as our starting point, and expect that the inheritance of routines signifies the birth of initial and lasting firm capabilities. Our fourth hypothesis is formulated as follows: Hypothesis 4: Spin-oﬀs of successful parents have lower hazards to exit, and this advantage over other entrants does not change as the firm ages. Thus far, we have addressed the issue of routine inheritance through spin-oﬀ dynamics and its eﬀect on firm survival. Successful incumbents are considered as superior learning environments and a source of succesful routines from aspiring entrepreneurs. However, other than generating more succesful spin-oﬀs, successful parent firms are also likely to generate higher numbers of spin-oﬀs than other parent firms. Better-performing firms are more likely to be a breeding ground of innovative and entrepreneurial eﬀort (Klepper, 2002; Klepper and Sleeper, 2005). Also, as a firm ages, it is likely to grow in size, and hence may have a higher amount of potential entrepreneurs. In addition, succesful firms and the success of their spin-oﬀs may contribute to a high reputation in the industry, which provides an edge on the labour market in their ability to employ the most creative and entrepreneurial young designers. By not only generating better-performing spin-oﬀs, but also doing so in greater numbers, successful incumbents fuel the diﬀusion of their routines in the industry population. To test this proposition we formulate the following hypothesis: Hypothesis 5: Firms are more likely to generate a spin-oﬀ at higher ages. Here, firm success is defined as the number of years a fashion house has been able to survive in the industry. Aside from firm age, we might expect the likelihood of generating a spin-oﬀ firm to be tied to the parent firm’s

The emergence of a firm genealogy in the fashion design industry

117

own background. Indeed, to the extent that pre-entry experiences constitute the transfer of organizational routines, we posit that parent firms with a spin-oﬀ or experienced background themselves are more likely to generate a spin-oﬀ, compared to parents which entered the industry as inexperienced entrants. We formulate our last hypothesis as follows: Hypothesis 6: Incumbents that were either spin-oﬀ or experienced entrants themselves, are more likely to generate a spin-oﬀ firm.

4.

DATA COLLECTION AND METHODOLOGY

4.1

Data Collection

Through our data collection we established a dataset of firm-level data on time of entry and exit, the complete career of the designer-founder and his or her main assistants, the firm’s location, and information on any mergers or acquisitions. We collected data from the onset of the industry in 1858 through 2005. The data are collected from accounts of the history of the fashion industry and biographies of its most prominent designers. The most important sources are Watson’s 20th Century Fashion: 100 Years of Style by Decade and Designer (2004) and the Thames & Hudson Dictionary of Fashion and Fashion Designers by O’Hara Callan (1998). The accounts were complemented by other, more specific biographers concerning particular couturiers (for example De Marly, 1980; De Rethy and Perreau, 2002; Sischy, 2004), and Internet sources, of which ‘www.designerhistory. com’ by Bawa (2007) has proven to be the most useful. In total, our dataset contains information on 565 of the world’s top fashion designers through the period 1858–2005. During our research period 510 designers started their own firm. This means that 55 designers have been active as employees only (up until 2005) and are left out of the analysis. Among them is Frenchman Gerard Pipart, who, although having had a successful career spanning 1949–1999 with successful organizations such as Balmain, Patou and Ricci, never started his own firm (Martin, 1997; Bawa, 2007). The industry continues to grow in terms of the number of firms (see Figure 6.5). While in most manufacturing industries, the number of firms first rises and then falls during an industry ‘shakeout’ (Klepper and Simons, 1997), this pattern is absent in the top fashion design industry. This is understandable because in creative activities, such as design, scale economies in production are less important in determining entry and survival, compared to manufacturing. Indeed, fashion designers are more concerned with remaining creatively relevant

Methods for analysing organizational routines

118

Number of fashion houses

400 350 300 250 200 150 100 50 1858 1867 1876 1885 1894 1903 1912 1921 1930 1939 1948 1957 1966 1975 1984 1993 2002

0

Time (years) Total population Source:

Entry

Exit

Own elaboration of O’Hara Callan (1998) and Watson (2004).

Figure 6.5

Number of top fashion houses, 1858–2005

and maintaining an image of luxury and exclusivity than with maximizing production to obtain low marginal costs (Banks et al., 2000). Considering that couture is a luxury good and the reliance of value on exclusivity, one could expect that an increase in demand does not necessarily translate to further growth of incumbents. Indeed, as the market grows, more and more niche spaces open up and these are as likely to be filled by new entrants as existing fashion houses. 4.2

Methodology

We test our hypotheses by estimating a Cox proportional hazard regression model. Through duration analysis, that is the number of years a firm has been able to survive in the fashion design industry, the model predicts the hazard to exit as conditioned by covariates of firm genealogy. We estimate hazards for fashion design firms to exit the market at time t dependent on n covariates x1, . . ., xn. The Cox proportional hazard function can be formulated as follows: h(t |x) 5 ␭0 (t)exp{␤1x1, . . ., ␤nxn}

(6.1)

Where h(t|x) is the hazard at time t dependent on covariates x 5 (x1, . . ., xn), ␤ 5 (␤1, . . ., ␤n) and ␭0 denotes the baseline hazard function.

The emergence of a firm genealogy in the fashion design industry

Table 6.2

119

Descriptive statistics of covariates in the Cox regression analysis

EXP.FIRM SPIN.FIRM YRS.PARENT. PROD

N

Minimum

Maximum

Mean

Std. Deviation

510 510 510

0.00 0.00 0.00

1.00 1.00 4.71

1.32

1.65

The duration model estimates the hazard to exit, or the time a fashion house is no longer active on the high fashion market. We counted the years between a firm’s entry and exit year as its spell duration, or ‘age’. The 340 firms that are still present in 2005 were treated as right-censored cases. Approximately 5 per cent of the total number of firms have at some point been acquired by other firms. We consider each acquisition by rival incumbent an exit event for the acquired firm. Table 6.2 shows descriptive statistics for each of the independent variables used in the Cox regression analysis. The first group of independent variables concerns the entrepreneurial background as earlier defined. We use a dummy for spin-oﬀ firms (SPINOFF) and a dummy for experienced firms (EXP.FIRM). Sometimes a firm had multiple founders, and whenever their backgrounds diﬀered, ‘spinoﬀ’ outranks ‘experienced’, and ‘experienced’ outranks ‘inexperienced’. In other words, we rank backgrounds by the relatedness of the pre-entry experience to the industry. In the few cases where multiple founders had spin-oﬀ backgrounds we compiled their preentry experiences at incumbents. We diﬀerentiate between spin-oﬀ firms by counting the maximum number of years a parent has produced at the time the employee leaves to start his own firm (YRS.PARENT.PROD). This number is used as a proxy for the success of the parent company. We derive these variables directly from our biographical data on each designer-founder. To reflect diminishing returns to scale in learning, we have transformed PARENTS and YRS. PARENT.PROD on a natural logarithmic scale. To test hypotheses 5 and 6, we estimate a logistic regression model to predict the probability of a spin-oﬀ event occurring based on parent firm characteristics. The logit regression function can be formulated as follows: Logit (Pi) 5 ln (Pi/1 2 Pi) 5 ␤0 1 ␤1x1,i 1 . . . 1 ␤n,ixn,i

(6.2)

Where the logits of the unknown binomial probabilities (P) of spin-oﬀ occurrance (event i) are modelled as a linear function of the variables x

120

Methods for analysing organizational routines

5 (x1, . . ., xn), ␤0 is a constant, and the parameters ␤ 5 (␤1, . . ., ␤n) are estimated through maximum likelihood. We analyse the probability of spin-oﬀ creation for all fashion design firms over the period 1925–2005. Because several parent firms have generated spin-oﬀs throughout their life course, we divide this research period in 16 five-year periods. The model derives for each time period per firm the probability that a spin-oﬀ event occurs based on (a) the number of years the firm is active prior to the spin-oﬀ event, and (b) entrepreneurial background. This measure allows us to examine at regular time intervals, the probability of spin-oﬀ creation based on a firm’s prior experiences. Several firms are active in more than one five-year period. In total, with an average firm age of 22 years, the number of cases accumulates to 2940 observations.

5.

RESULTS

We tested diﬀerent hazard models using Cox regression with an increasing number of variables, shown in Table 6.3. We start with Model 1 where we include entrepreneurial background only. Following hypothesis 1 and hypothesis 2, spin-oﬀ firms and experienced firms indeed have lower hazard than the omitted category of inexperienced start-ups. Both coeﬃcients are negative and significant. Thus, we accept hypothesis 1 and 2. Figure 6.6 shows the survival graphs of the population of firms per background. It shows the share of firms that survive up to a particular age. Spin-oﬀ and experienced entrants clearly outperform other entrants in terms of survival. Model 2 replaces the spin-oﬀ variable with the number of years the parent firm has been active in the industry at the time of spin-oﬀ formation. In doing so, we take into account whether an entrepreneur has previous experience in working at an incumbent, and diﬀerentiate between spin-oﬀs based on quality of parentage. The coeﬃcient for YRS. PARENT.PROD is negative and significant, indicating that spin-oﬀs with successful parents outperform spin-oﬀs with less successful parents. This result suggests that (parts of) organizational routines and practices are transferred from parents to their spin-oﬀs. We accept hypothesis 3. Figure 6.7 shows the survival graphs of the population of spin-oﬀ firms per parent age category. It shows the share of spin-oﬀs that survive up unto a particular age and does by age category of the parent firm. The higher categories of parent age (31 years and higher, and 11 to 31 years) constitute 20 per cent each of the population. These firms were formed by employees from firms that survived for over 10 years. When we compare

The emergence of a firm genealogy in the fashion design industry

Table 6.3

Estimates of the Cox regression models (standard errors are mentioned in parentheses; dependent variable: hazard to exit)

EXP.FIRM SPIN.FIRM

Model 1

Model 2

Model 3

21.388*** (0.219) 21.252*** (0.225)

21.248*** (0.208)

21.199*** (0.209)

20.363***

20.549***

(0.070)

(0.128) 0.006*

YRS.PARENT. PROD YRS.PARENT. PROD*t Number of cases (N) Chi-square 22 Log likelihood

510 48.919*** 1685.948

(0.004) 510 35.682*** 1683.709

510 32.442*** 1686.949

Notes: *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level.

Background Spin-off

1.0

Experienced 0.8

Inexperienced

Cum Survival

0.6

0.4

0.2

0.0 0

Figure 6.6

121

10

20

30 Age

40

50

60

Survival graph of the fashion design firm population by preentry background

Methods for analysing organizational routines

122

Spin-off background Parent age 31 years or more Parent age 11 to 31 years Parent age below 11 years

1.0 0.9

Cum Survival

0.8 0.7 0.6 0.5 0.4 0.3 0

Figure 6.7

10

20

30 Age

40

50

60

Survival graph of the spin-oﬀ firm population by parent age

Figures 6.6 and 6.7, it becomes clear that even though on average all spinoﬀs outperfom inexperienced entrants, spin-oﬀs from successful parents outperform other spin-oﬀs. In Model 3 we test hypothesis 4, regarding the extent to which the inheritance eﬀect of YRS.PARENT.PROD on firm survival lasts throughout a spin-oﬀ’s life time. In line with Klepper’s model of industry evolution, we posit that this eﬀect should be proportional to all ages. In Model 3 we add the variable YRS.PARENT.PROD*t, that is the time-dependent covariate of YRS.PARENT.PROD. Since we expect the hazard eﬀect of routine inheritance to be proportional, we should not see a significant eﬀect of the time-dependent variable. Interestingly, the coeﬃcient of YRS.PARENT. PROD*t is positive and significant. The other coeﬃcients remain similar to Model 2. This result indicates that the negative eﬀect of YRS.PARENT. PROD on the hazard rate diminishes with firm age. We reject hypothesis 4. Hypotheses 5 and 6 concern the probability of spin-oﬀ creation by parent firms. Successful parent firms are posited as more fertile learning environments for organizational routines for entrepreneurial employees. Hence, we expect firms that have accumulated learning-by-doing experiences over time and that were able to survive longer to be more likely to generate spin-oﬀs. Indeed, in Model 1 in Table 6.4, we can see that the coeﬃcient of FIRM.AGE, that is the number of years the firm exists, positively aﬀects its probability to generate a spin-oﬀ company. More successful firms, in terms of survival, have a higher chance of creating spin-oﬀs. We accept hypothesis 5. Interestingly, in Figure 6.8 we can see that most

The emergence of a firm genealogy in the fashion design industry

Table 6.4

123

Estimates of the logistic regression models (standard errors are mentioned in parentheses; dependent variable: probability to generate a spin-oﬀ)

FIRM.AGE EXP.FIRM SPIN.FIRM

Model 1

Model 2

1.094*** (0.061) 1.255*** (0.267) 1.401*** (0.269)

1.151*** (0.062) 1.067*** (0.175)

YRS.PARENT.PROD Constant Number of cases (N) Chi-square 22 Log likelihood

25.372*** (0.308) 2940 597.357*** 2704.772

1.238*** (0.163) 25.360*** (0.258) 2940 624.176*** 2677.953

Notes: *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level.

spin-oﬀs are created after the parent firms have been active for about ten years in the industry. Figure 6.8 also shows that the number of spin-oﬀs generated is dependent on background of the firm. Spin-oﬀ firms and experienced firms seem to generate the bulk of spin-oﬀs in the fashion design industry. This is confirmed by the positive and significant coeﬃcients of EXP.FIRM and SPIN.FIRM in Model 1 of Table 6.4. We accept hypothesis 6. These results confirm the importance of spin-oﬀ dynamics in the industry. Experienced firms entered early and outperformed inexperienced entrants. Since the nineteenth century the master–pupil system of couture translates into a spin-oﬀ dynamic which replicates and diﬀuses routines to new generations of design houses. These spin-oﬀs contribute to the process as they themselves pass on organizational routines to future generations of spin-oﬀs. Our analysis shows that spin-oﬀs are likely to create spin-oﬀs themselves. In Model 2 of Table 6.4 we show that spin-oﬀs from better-performing parents are more likely to generate spin-oﬀs themselves. Unfortunately, and similar to our survival analysis, the multicollinearity problem does not allow for both SPIN.FIRM and YRS.PARENT.PROD to be considered

Methods for analysing organizational routines

124

Number of spin-off firms generated

14

Parent with ‘spin-off’ background

12 10

Parent with ‘experienced’ background

8 6 4

Parent with ‘inexperienced’ background

2 0 1

Source:

8

15 22 29 36 43 50 57 Age (in years)

Own elaboration of O’Hara Callan (1998) and Watson (2004).

Figure 6.8

Number of spin-oﬀs generated by parent firms, per age and background of the parent

in one model. The coeﬃcient of YRS.PARENT.PROD is positive and significant, indicating that spin-oﬀs that seem to have inherited ‘fitter’ routines, are more likely to generate new spin-oﬀs.

6.

CONCLUSIONS AND DISCUSSION

In this chapter we have addressed the question regarding whether organizational routines are transferred between couture firms, and to what extent this may contribute to firm survival and on the aggregate level to the evolution of routine heterogeneity in a firm population. Head designers and assistant designers are proposed here to engange in a mentor–student type of relatonship in which not only technical skill, but also organizational routines are transferred from master to pupil, from parent firm to spin-oﬀ company. We found that experienced firms and spin-oﬀs outperformed inexperienced entrants, indicating that related pre-entry experience benefits entrepreneurs in their endeavours. More importantly, our results indicate that spin-oﬀs inherit their parent’s organizational routines, albeit imperfectly. We construct the genealogy of the high fashion design industry based on the idea of routine inheritance, and find that most design houses of today can trace their roots to a handful of early experienced entrants. These firms generated spin-oﬀs, which inherited their successful routines, and created spin-oﬀs themselves. Successful parents created successful spin-oﬀs. In addition, of course, this result may be restricted to fashion,

The emergence of a firm genealogy in the fashion design industry

125

considering this industry’s focus on short product life cycles. Indeed, many designers draw on past experiences for inspiration, and many elements of past designs are seen on present day catwalk shows, reincorporated in new collections. Fashion houses function as the seedbeds of future entrepreneurs who learned of organizational practices while working for these established firms. It seems that the concept of organizational routine can be applied just as well within a creative industry as it has been in various other industries, such as automobiles, lasers and television manfacturers. By approximating organizational routines by the pre-entry experiences of the founder, we are able to explain the individual success of fashion houses throughout the industry’s history. In this respect, our results strengthen the generality of the evolutionary theory of routine inheritance through spin-oﬀ dynamics. A shortcoming of our analysis, however, is that we made no in-depth inquiry into fashion designers’ work-processes. Routines are unobserved in this chapter and we do not measure routine replication as such, aside from a few anecdotes on exemplary cases. We attempted to measure the impact of routine replication, and treated the organizational routine as a black box. As Pentland and Feldman (2005) point out, the use of routines as a black box is useful for these kinds of analysis but remains inaccurate. Indeed, more fine-grained empirical research is needed to validate our assumptions. Note that the same holds for many similar studies of parent– progeny routine transfer (for example Phillips, 2002; Wezel et al., 2006; Klepper, 2002). Reconciling routinization of work practices and creative behaviour remains one of the literature’s most important areas for future research. The importance of gaining experience as assistant of an established fashion designer reflects that creativity is not purely a personal trait, but can be transmitted if two people work closely together. Our results show that the spin-oﬀ mechanism of routine diﬀusion has influenced firm performance on the micro-level, and market structure on the meso-level. Successful parents produce better, and more, spin-oﬀs. These spin-oﬀs, in turn, pass on their fitter routines to their progeny. As a fitter routine spreads, however, competition ensures that entry possibilities for less fit routines decrease. As such, routine transfer can explain the success of individual founders, and of accompanied higher entry barriers in the industry. More and more entrants will find the need to learn from incumbents before they can make a successful start of their own. Our results suggest that spin-oﬀ generation acts as a mechanism of routine diﬀusion and culminates in a genealogy which can explain firm survival and, on the aggregate level, the market structure of the industry.

126

Methods for analysing organizational routines

Spin-oﬀs outperform other entrants, and better-performing parents generate better-performing spin-oﬀs. However, at least for the fashion design industry in our research period, this inheritance eﬀect diminishes over time. The organizational blueprints which have been set up by the entrepreneur seem to be based on his or her pre-entry experiences, and have a significant eﬀect on the firm’s subsequent survival. However, as the new firm ages, the initial eﬀect of routine inheritance diminishes. This is suggestive of change in organizational routines laid down by the designer-entrepreneur. Indeed, firm capabilities may be born from the pre-entry experiences of the firm’s founder, but these capabilities are complemented over time by experiences of new personnel and eventually the replacement of the founder as head designer. Our results are an indication of post-entry dynamic of adaptation of organizational routines to circumstances. These adaptations may build upon the initial foundations handed down from the entrepreneur’s pre-entry experiences, yet our results are suggestive of post-entry learning dynamics not yet addressed in the spin-oﬀ model of industry evolution. To what extent do spin-oﬀs have an edge in adapting to changing circumstances? What eﬀect does the in- and outflow of key personnel have on the construction or deterioration of the existing organizational routine structure? If the inheritance eﬀect of organizational routines diminishes over time, how are spin-oﬀ firms able to keep their competitive edge? These questions form new challenges as part of an exciting evolutionary agenda which aims to complement existing theories on the replication and diﬀusion of organizational routines.

REFERENCES Agarwal, R., R. Echambadi, A.M. Franco and M.B. Sarkar (2004), ‘Knowledge transfer through inheritance: spinout generation, development, and survival’, Academy of Management Journal, 47(4), 501–22. Banks, M., A. Lovatt, J. O’Connor and C. Raﬀo (2000), ‘Risk and trust in the cultural industries’, Geoforum, 31, 453–64. Bawa, P. (2007), ‘History of fashion and costume’, available at www.designer history.com. Becker, M.C. (2004), ‘Organizational routines. A review of the literature’, Industrial and Corporate Change, 29, 249–62. Boschma, R.A. and R. Wenting (2007), ‘The spatial evolution of the British automobile industry’, Industrial and Corporate Change, 16(2), 213–38. Buenstorf, G. (2006), ‘How useful is generalized Darwinism as a framework to study competition and industrial evolution?’, Journal of Evolutionary Economics, 16, 511–27. Burton, M.D., J.B Sorensen and C. Beckman (2002), ‘Coming from good stock: career histories and new venture formation’, Research in the Sociology of Organisations, 19, 229–62.

The emergence of a firm genealogy in the fashion design industry

127

Cantner, U., K. Dressler and J.J. Krüger (2006), ‘Firm survival in the German automobile industry’, Empirica, 33, 49–60. Chapon, F. (1996), Jacques Doucet – Ou l’art du mécénat, 1853–1929, Paris: Perrin. Dahl, M.S. and T. Reichstein (2005), ‘Are you experienced? Prior experience and the survival of new organisations’, DRUID Working Papers 05-01, www.druid.dk. Dahl, M.S., C.O.R. Pedersen and B. Dalum (2003), ‘Entry by spinoﬀ in a high-tech cluster’, DRUID Working Papers 03-11, www.druid.dk. De Marly, D. (1980), The History of Haute Couture 1850–1950, New York USA: Holmes & Meier Publishers. De Rethy, E. and J.L. Perreau (2002), Christian Dior: The Early Years 1947–57, New York, USA: Vendome Press. Florida, R. (2002), The Rise of the Creative Class, New York USA: Basic Books. Ford, C.M. and D.A. Gioia (2000), ‘Factors influencing creativity in the domain of managerial decision making’, Journal of Management, 26, 705–32. Gort, M. and S. Klepper (1982), ‘Time paths in the diﬀusion of product innovations’, The Economist, 92, 630–53. Helfat, C.E. and M.B. Lieberman (2002), ‘The birth of capabilities: market entry and the importance of pre-history’, Industrial and Corporate Change, 11, 725–60. Hodgson, G. and T. Knudsen (2004), ‘The firm as an interactor: firms as vehicles for habits and routines’, Journal of Evolutionary Economics, 14, 281–307. Klepper, S. (1996), ‘Entry, exit, growth, and innovation over the product life cycle’, American Economic Review, 86, 562–83. Klepper, S. (1997), ‘Industry lifecycles’, Industrial and Corporate Change, 6, 145–82. Klepper, S. (2002), ‘The capabilities of new firms and the evolution of the US automobile industry’, Industrial and Corporate Change, 11(4), 645–66. Klepper, S. (2007), ‘Disagreements, spinoﬀs, and the evolution of Detroit as the capital of the U.S. automobile industry’, Managament Science, 53, 616–31. Klepper, S. and K.L. Simons (1997), ‘Technological extinctions of industrial firms: an inquiry into their nature and causes’, Industrial and Corporate Change, 6, 379–460. Klepper, S. and K.L. Simons (2000a), ‘The making of an oligopoly: firm survival and technological change in the evolution of the US Tyre Industry’, Journal of Political Economy, 108(4), 728–60. Klepper, S. and K.L. Simons (2000b), ‘Dominance by birthright: entry of prior radio producers and competitive ramifications in the US television receiver industry’, Strategic Management Journal, 21(10–11), 997–1016. Klepper, S. and S. Sleeper (2005), ‘Entry by spinoﬀs’, Management Science, 51(8), 1291–306. Kloosterman, R.C. (2004), ‘Recent employment trends in the cultural industries in Amsterdam, Rotterdam, The Hague and Utrecht: a first exploration’, Tijdschrift voor Economische en Sociale Geografie, 95(2), 243–52. Maddison (2007), ‘Historical statistics for the world economy: 1–2003 AD’, available at www.ggdc.net/maddison. Martin, R. (1997), The St. James Fashion Encyclopedia: A Survey of Style from 1945 to the Present, Canton, MI, USA: Visible Ink Press. Merlo, E. and F. Polese (2006), ‘Turning fashion into business: the emergence of Milan as an international fashion hub’, Business History Review, 80(3), 415–47.

128

Methods for analysing organizational routines

Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Harvard University Press. O’Hara Callan, G. (1998), The Thames & Hudson Dictionary of Fashion and Fashion Designers, London, UK: Harry N. Abrams. Ohly, S., S. Sonnentag and F. Pluntke (2006), ‘Routinization, work characteristics and their relationships with creative and proactive behaviours’, Journal of Organizational Behaviour, 27(3), 257–79. Pentland, B.T. and M.S. Feldman (2005), ‘Organizational routines as a unit of analysis’, Industrial and Corporate Change, 14(5), 793–815. Phillips, D.J. (2002), ‘A genealogical approach to organizational life chances: the parent–progeny transfer among Silicon Valley law firms, 1946–1996’, Administrative Science Quarterly, 47(3), 474–506. Porter, M. (2000), The Comparative Advantage of Nations, New York, USA: Free Press. Scott, A.J. (2000), The Cultural Economy of Cities, London, UK: Sage Publications. Scott, S.G. and R.A. Bruce (1994), ‘Determinants of innovative behavior: a path model of individual innovation in the workplace’, Academy of Management Journal, 37, 580–607. Sischy, I. (2004), The Journey of a Woman: 20 Years of Donna Karan, New York, USA: Assouline. Stinchcombe, A.L. (1965), ‘Social structure and organisations’, in J.G. March (ed.), Handbook of Organizations, Chicago, IL, USA: Rand McNally & Company, pp. 142– 93. Teece, D., G. Pisano and A. Shuen (1997), ‘Dynamic capabilities and strategic management’, Strategic Management Journal, 18(7), 509–33. Vinodrai, T. (2006), ‘Reproducing Toronto’s design ecology: career paths, intermediaries, and local labor markets’, Economic Geography, 82, 237–64. Waddell, G. (2004), How Fashion Works: Couture, Ready-to-Wear and Mass Production, New York, USA: Blackwell Publishing. Watson, L. (2004), 20th Century Fashion: 100 Years of Style by Decade and Designer, in Association with Vogue, New York, USA: Firefly Books. Weick, K.E. (1998), ‘Improvisation as a mindset for organizational analysis’, Organization Science, 9(5), 543–55. Wenting, R. (2008), ‘Spinoﬀ dynamics and the spatial formation of the fashion design industry 1858–2005’, Journal of Economic Geography, 8(5), 593–614. Wezel, F.C., G. Cattani and J.M. Pennings (2006), ‘Competitive implications of interfirm mobility’, Organization Science, 17(6), 691–709. Zellmer-Bruhn, M.E. (1999), ‘The eﬀects of time pressure and interruptions on team external acquisition of work routines’, PhD dissertation, University of Wisconsin, Madison, WI, USA.

PART III

Organizational routines and stability in organizations

7.

Is it the ‘same’? Observing the regeneration of organizational character at Camp Poplar Grove1 Jeremy P. Birnholtz, Michael D. Cohen and Susannah V. Hoch

1.

THE MIRACLE OF ORGANIZATIONAL REGENERATION

‘Organizational regeneration’, a phenomenon most clearly visible in seasonal organizations, such as summer camps and ski areas, is the process of reproducing an organization after a period of dormancy, often by training and guiding personnel who are largely new to their roles, or to the organization as a whole. This process presents a revealing instance of a fundamental problem facing all members of organizations and those who study them: how can we talk of an organization as being the same entity over time? How can we attribute properties today to a business, government agency, or school based on observations made yesterday? After nine months of inactivity, an established summer camp ‘comes back to life’. In general, campers, parents and staﬀ, while perhaps noticing some distinctive features of this year’s incarnation, regard the camp as the same one they have known from previous years. This is all too easily taken for granted. But looking carefully convinces us that it is a kind of small miracle, an occasion for wonder much like what one finds in the garden each spring. Our intent is to understand how such regeneration is possible. We believe that in doing so we can shed light on how groups develop and maintain the collective property of ‘being organized’. The Paradox of the [N]ever Changing World Many have argued (for example, Cohen, 1999; Pentland and Feldman, 2002) that when studying recurring action patterns in organizations, such as routines, there is substantial truth in both of these proverbs: 131

132 ● ●

Organizational routines and stability in organizations

one does not step into the same river twice (Heraclitus); there is no new thing under the sun (Ecclesiastes).

From one perspective (what might be called extremely close up), no situation of action repeats itself identically. From an alternative perspective (what might be called standing far back), every action appears as but a variant on, or recombination of, some predecessors or known general types of actions. This ‘paradox’ of the (n)ever-changing world may be more or less apparent, but it never really vanishes. Always in its shadow, an organization must synthesize the diverse, typically inconsistent, capabilities and preferences of its members into a coherent ecology of recurring actions that aﬀects the world in a recognizable way. To continue receiving the resources it requires, it must be able to aﬀect the world at a level exceeding some minimal competence. Indeed, if that loop is not closed we should hesitate to call the system organized. In doing so, organizations must convert perpetual novelty into actionable similarity (Weick, 1995). At the same time, they must also avoid mistaking significantly novel conditions as occasions for mere repetition of a familiar response. The former issue is particularly acute in the conditions of our chosen case. In seasonal organizations such as Camp Poplar Grove, which is described in detail below, activity is interrupted for long periods, and most of the assembled actors for the next cycle are inexperienced in their roles or are completely new to the organization. As they gather to (re) create a summer camp together, very many of the ‘things under the sun’ seem new, and appropriate actions are therefore not obvious. Still, the case of Poplar Grove clearly illustrates that a few weeks suﬃced to regenerate a complex system of interdependent activity (what we term an ecology) that is recognizably another instance of ‘the same’ organization. The term ‘flagpole’, for example, was quickly understood by camp newcomers to refer to both a location (for example, ‘Meet me at flagpole’) and an all-camp assembly (for example, ‘We’ll have flagpole at 10:45’), despite the flag and pole having both been removed years before. We claim that at Camp Poplar Grove there is rapid regeneration of many collective action patterns when only a few of the actors have direct prior experience. The claim that eﬀective action patterns are generated is substantiated by the evident functioning of the Camp: it takes in hundreds of campers, guides them through weeks of satisfying activities, and sends them safely home. The claim that after the first year, despite dormancy and turnover, this has been re-generation is more subtle and requires some careful distinctions. Re-generation implies sameness. In what sense is the Camp ‘the same’ year after year? Clearly, parents and campers believe it to be eﬀectively the same. A child, and then his or her siblings, will often be in residence

The regeneration of organizational character at Camp Poplar Grove

133

for a decade or more in large part because campers and parents expect the Camp to oﬀer a similar and valued experience. Eisner (2005), for example, discusses the similarity of camp experiences at a Vermont camp for three generations of his family. Such experiences strongly suggest that this continuity of commitment is achieved by the actions of the Camp’s personnel, and is not produced by simple physical circumstances. The mere existence of a set of buildings by a lake along with equipment for a program of activities does not suﬃce to guarantee a consistent experience. Although continuity of clientele and recognition by alumni testify to some level of achieved similarity, the Camp cannot be identical yearto-year. In the Heraclitean extreme, nothing is identical, and certainly a summer camp, with heavy turnover in counselors and staﬀ – not to mention variations in campers and the vagaries of weather – cannot be exactly the same in any two years. In this chapter we draw on a detailed examination of one cycle of regeneration at Camp Poplar Grove and develop the notion of ‘organizational character’ as an aid to understanding the ability of the participants to regenerate what we will describe as a coherent ecology of action patterns that is recognizable as another instance of ‘the same’ organization. We will argue that this coherent set of action patterns is recognizable as organizational character, and that it is passed on via demonstration, rapid bursts of communication, and the iterative application of generic skills in specific contexts. Note that when we refer to coherent ecologies of action dispositions, we deliberately avoid the term ‘routine’, except for brief mentions in our conclusions. While it is certainly possible that some of the action patterns we describe may form the core of the sequential, durable sequences of action that are often labeled as routines, we use more general terms here to avoid definitional issues and the connotations of mindlessness that often surround the notion of routine (Cohen, 2007; Cohen et al., 1996). This chapter is unusual in its structure, in order to spark discussion and invite alternative interpretations of our data. In the next section we present a summary of our case data, followed by a discussion on the notion of organizational character and observations on regenerative processes observed in the Poplar Grove case.

2.

OBSERVING REGENERATION AT POPLAR GROVE

Our discussion is based on experience with and observation of Camp Poplar Grove (a pseudonym), a private summer camp in northern Michigan. Poplar Grove was founded in 1955 to foster independent thinking and

134

Organizational routines and stability in organizations

decision making in children, and continues to be owned and operated by the founders’ son, Richard (a pseudonym, as are all subsequent proper names except Jeremy Birnholtz and Susannah Hoch), who gradually took control starting in 1975. The Camp’s facility, shown in Figure 7.1, spans about 400 acres, and includes facilities for various activities, a small private lake, cabins, an infirmary, and a main lodge with kitchen and dining facilities. There are 70–100 campers at camp during each of four two-week sessions and the Camp employs 25–35 service and program staﬀ members each summer. Turnover in personnel is high, as is common in summer camps. Of the 33 staﬀ and administrators we observed, 23 were new to the Camp, and 18 new to camp counseling entirely. During the rest of the year the directors, Richard and his wife Michelle, handle administrative operations, and the maintenance director works part-time at Camp in the spring and fall. The Camp otherwise has no year-round staﬀ. Data Gathering and Analysis The primary data for this study were gathered via minimally obtrusive participant observation carried out over 22 days in 2003 by Susannah Hoch. This period spanned the Camp’s staﬀ training week and the first two weeks of its operation with campers. Hoch was unfamiliar with the Camp and comparable in age and background to the counselors being observed. She worked from guidelines that were as minimal as possible, so as to reduce expectation biasing. She was asked to observe how counselors came to understand and carry out their roles and responsibilities at Camp. She was also invited to participate in training and staﬀ activities, and to reflect on her own experience as a newcomer to the Camp. She did not identify Birnholtz, who was on-site as assistant director, as having a research interest in her work, and she did not share results with him until after he completed the year’s work with the Camp. Hoch’s field notes (Emerson et al., 1995) were generally produced in three to five short writing sessions per day, yielding 72 field note documents, ranging in length from 191 to 2066 words, with a median of 758. Near the end of the observational period, Hoch also conducted semistructured interviews with eight staﬀ members. One of us (Jeremy Birnholtz) also has a long personal association with Poplar Grove, over a period spanning 18 years and involving 11 seasons in residence as a camper, counselor, and eventually as assistant director of the Camp. We draw on Birnholtz’s experience largely in providing historical context for and in illustrating details observed by Hoch.

The regeneration of organizational character at Camp Poplar Grove

Figure 7.1

Sketch map of Camp Poplar Grove

135

136

Organizational routines and stability in organizations

Data analysis consisted of independent reading and re-reading of Hoch’s field notes. Episodes of interest were identified and tracked, with a particular focus on three types of circumstances: 1) apparent uncertainty on the part of an individual about how to proceed and subsequent action; 2) corrections to ‘incorrect’ actions taken by individuals; and 3) innovative actions that appeared to take hold or be extinguished. While it was sometimes diﬃcult to infer from field notes when individuals were uncertain about how to act or what drove them to act in particular ways, we believe we were able to reasonably extrapolate this information from the combination of Hoch’s records of manifest uncertainty, her detailed reflections on her own experience as a newcomer to the Camp, her informal conversations with staﬀ members throughout her observation period, and notes from reflective interviews with staﬀ. Summary of Observations In this section, we present a short summary of Hoch’s observations which is intended to orient the reader to the general setting and sequence of events. Field observations at Poplar Grove began on Friday, 13 June, 2003, two days before the staﬀ ‘training week’. Physical reopening of the Camp from its winter shutdown was well underway, but the major eﬀort accompanying the arrival of the first summer staﬀ was just starting. Major tasks for these first days included bringing the kitchen into service to feed the arriving staﬀ, and general maintenance work such as cutting grass and painting. Hoch initially spent one day helping out and observing in each of these areas. Staﬀ at Camp at this point included Jeremy, the assistant director; Michelle and Richard, the directors; Rick, a former head cook and counselor who was helping with the opening; and Lisa, the maintenance assistant who had arrived four days earlier and was new to the camp. With the exception of Lisa, everyone at Camp so far had experience in all its areas, and drew on these experiences in guiding Hoch and in making their own decisions. Receiving guidance from these people raised for Hoch issues that would be seen many times among the counselors: whose advice to follow when conflicting was guidance oﬀered; whose advice to seek when uncertain; and whether to improvise when there was doubt or to seek further guidance before acting (that is, ‘whether to ask forgiveness or permission’). Another interesting aspect of these first few days was the ‘one-eyedamong-the-blind’ dynamic that Hoch observed. Specifically, she noticed that Lisa, despite having only been at Camp for four more days than the

The regeneration of organizational character at Camp Poplar Grove

Figure 7.2

137

‘Flagpole’ with campers and counselors

other staﬀ, was someone to whom new staﬀ (and Hoch herself) regularly turned when they had questions about where things were kept and how things were done. Thus, small diﬀerences in experience were amplified, making experts of those with only modest knowledge advantages, a process that gives rise to what we’ll later call a cascade. On the afternoon of the second day of observations, the new counselors began to arrive. The early arrivals and those experienced at Poplar Grove began filling in basic orientation for the others and making plans for the upcoming day. On the third morning, 15 June, Hoch went to observe the kitchen and to meet Cathy, the new cook. Witnessing the first of many minor breakdowns, she found breakfast ready as Cathy had been told the staﬀ – now numbering 24 – would come at 9:00. However, the staﬀ believed breakfast time was 9:30, and were gathered some distance away in the area called ‘flagpole’, the area that continued to serve as the regular assembly site before each meal (see Figure 7.2). Over breakfast, there were casual conversations about prior camp experience at Poplar Grove and elsewhere, and about a general orientation provided to some new counselors by an international recruiting agency involved in hiring some of the staﬀ. Jeremy announced a schedule of the day’s activities, including mealtimes, setting up of facilities, and staﬀ meetings. Post-breakfast was followed by a period of tasks getting underway. This was a halting process as there was much uncertainty about who should do

138

Organizational routines and stability in organizations

what, and a shortage of explicit direction. Counselors with prior experience, and Lisa, began giving direction and recruiting volunteers for various tasks. Much work was centered on preparation of the waterfront and sailboats. Many staﬀ were engaged in unfamiliar tasks and asked each other for guidance (‘how do I?’, ‘should I. . .?’) or for assessment of whether a task had been done correctly (‘Is this right?’). Before lunch Hoch returned to the kitchen, where there were many questions about how exactly the eating space should be set up to handle what would be a large flow of eaters when 75 campers arrived the following week. Cathy’s eﬀorts to solve this problem were ambivalent. She was an experienced cook at resorts and camps, and appreciated that each has its own methods. This she expressed by saying ‘I am sure they have a system that already works, so why should I invent a new system?’ At the same time, she did have ideas about common problems of traﬃc flow and buﬀet table set-up, and so did have definite preferences about some aspects of the layout. Over the course of the observations, this tension would be evident, within her and between her and other experienced Camp staﬀ, such as Michelle. Hoch observed Cathy making records of her kitchen management decisions that would useful to her if she returned, or to a successor. No such records had been inherited from previous occupants of the cook role. As the week went on, the kitchen began to function a little more smoothly in serving the staﬀ. However, some of the work was still aided by Rick and Michelle taking part and demonstrating established practices, along with supervision from Cathy. The ‘girls’ (as Cathy called them) who formed the kitchen staﬀ – three from Poland, one from England2 – now had a clear sense of the basic mealtime procedures, and guided others who did things incorrectly. Cathy used an on-the-fly quiz approach to correcting minor discrepancies: ‘oh, you forgot something. . .’ followed by a pause while the staﬀ member figured out what was missing. There were also minor, though good-natured, tugs of war between Cathy and Rick and Michelle about arranging serving stations to accommodate flow. On day four, one major job was putting the dock into the lake. The new staﬀ didn’t have a useful overview of the eventual set-up, and there was discussion of whether ‘like an “H”’ meant lower or upper case. Pictures were etched in the sand, and Richard, the Camp director, came to the beach and gave a short orienting talk about the overall eﬀort and the division of labor. Over the course of the job, he explained what the next episode would be in the task. At the end, with the pieces in their correct places, one of the counselors remarked that perhaps they should be marked so that future deployments would be easier. While this was acknowledged to be a good

The regeneration of organizational character at Camp Poplar Grove

139

suggestion, nobody marked the pieces as the counselors hurried oﬀ to do other jobs or change into dry clothes. The predominant activities for the rest of training week were preparation of cabins, periodic group meetings to review rules and regulations, and the planning and practicing for instruction in the Camp’s activity areas, such as water skiing, drama, tennis, horseback riding, and others. Each of these areas would be staﬀed by one or two of the counselors, who – for the most part – began considering lesson plans and asked questions about materials and procedures, often of experienced counselors, but sometimes of Jeremy or, less often, of Richard. There were also group meetings that reviewed rules written in manuals, but which seemed to garner counselor attention principally through oral presentation. These meetings also led to restless boredom after a while, with the notable exception of water safety discussions led by Richard on the topic of how to deal with a possible drowning. This did seize attention. Hoch noticed over the days of setting up that a pride of ownership developed, seeded from the returning staﬀ, but also noticeable in the new staﬀ. This atmosphere supported persistence in learning jobs, and patience in lending others a hand and answering their questions. Another part of the scaﬀolding for organizing the Camp was laid down in meetings that were led by Jeremy or Richard that discussed overriding goals such as ‘create an environment for the kids to have fun and be safe’. Though not especially precise, these statements conveyed priorities that were useful in later, distributed episodes of reasoning about dilemmas that arose. By the end of the seven-day training period it was apparent that some staﬀ more consistently knew what was going on. This may have been partly a function of personality, and may also have been related to work roles aﬀording observation, such as oﬃce assistant. Whatever the bases, these people were recognized as better sources for information about plans and problems. On the tenth day of observations, the campers arrived. Their presence placed a new set of stresses on the practices developed so far by the staﬀ. Counselors gathered in the morning for flagpole and were divided into teams of greeters, unpackers, and game players to cover requirements of moving in 75 campers over the course of the day. Hoch’s observations now centered more on the cabins for which the counselors were responsible, as the arrival of the campers raised many issues of maintaining physical facilities (such as unplugging toilets) and finding workable strategies for flexibly enforcing rules on campers (for example, bedtimes and clean-up responsibilities). This process of establishing the full functioning of the Camp was strongly aﬀected by returning campers, who

140

Organizational routines and stability in organizations

deployed their extra experience both to gain some advantages for themselves, but also – and much more commonly – to help new campers and staﬀ learn the patterns of a Poplar Grove summer. Counselors confronted new situations, and this renewed jokes among them about rules that were covered in the manuals but hadn’t been read, or were misconstrued, not remembered, or bypassed. The afternoon of the tenth observation day, 22 June, was dominated by initial activities for the arrived campers. First were waterfront orientation and swim tests. These introduced the campers to rules about the Camp’s buddy system for swimming in pairs, and its chip system that was used to indicate campers’ qualification level for waterfront activities, and to track who was in the water at any given time. Richard, as director, made a brief talk stressing the import of safety and respecting the authority of Amy and Laura, the head lifeguards. By this point Hoch could report that staﬀ were quite comfortable with her observer role. She had found ways to describe her research as sponsored by her university without ever having to mention that Jeremy would be part of the analysis team once the summer ended. As the tenth and eleventh observational days unfolded, Hoch visited many diﬀerent parts of the Camp to watch the counselors establish the practices, rules and rhythms of their activity areas. Horseback riding, like the waterfront, presents substantial safety concerns – for both campers and horses – and therefore involves substantial pressure on the counselors to establish rules. At the other extreme, many crafts projects can be pursued quite independently and require counselors only for intermittent advice about materials and methods. During these days, both campers and counselors worked into the system of five activity periods per day. They learned the schedules of what they were to be doing at each time of the day, and learned to use the information board that provided schedule data. This settling-in period contained a number of minor breakdowns that became occasions for further learning. One involved the waterfront chip system. Campers who were solid swimmers got a white chip. Red chip campers were non-swimmers, confined to the shallows. Blue chip holders had intermediate privileges that didn’t include sailboats or water skiing. The campers left their chips with the waterfront supervisors when they were out in the water and picked up from a bucket when they left the area, allowing the supervisors to know if anyone was unaccounted for at the waterfront. The diﬃculties arose when waterfront counselors issued ‘conditional white chips’ to some campers who were to have one more swimming lesson before having full privileges. Procedures of other counselors then broke down as the holders of full and conditional white chips were

The regeneration of organizational character at Camp Poplar Grove

141

indistinguishable although they were intended to have diﬀerent privileges. Investigating the origins of the ‘innovation’ and resolving the diﬃculties drew in many members of the staﬀ and constituted a major learning episode for the organization. The conditional white chip turned out to be an in-the-moment invention of Richard’s, meant to deal with campers who appeared to be skilled enough, but were resting during their required six-lap swim. This innovation was extinguished – no ‘conditional white chips’ were issued by the waterfront staﬀ during swim tests in the subsequent three sessions of camp. Another breakdown occurred on day 15, when a counselor assigned kitchen duty to members of his cabin as a punishment for sleeping in too late. But this caused problems for Cathy who then had to teach kitchen jobs to inexperienced helpers. The innovation was also extinguished. Observation day 13 involved a day trip, and day 14 was marred by rain. This required shifting scheduled outdoor activities to indoor substitutes. Considerable confusion ensued as unpracticed fallback activities and locations were deployed. At one point Kelly, a returning counselor, met a group of new counselors and gave them very definite instructions: ‘Go to your second activity and see if there are campers there. You must make sure you can account for all of them. You can have your activity outside if you want, but just make sure you know where all your campers are if they are in arts and crafts or something’. The evening was devoted to a carnival, in which every cabin provided a booth that oﬀered some kind of game or entertainment. Brian, a counselor with several years of Poplar Grove experience as camper and training staﬀ, was in charge of the event and worked hard to re-implement practices from earlier years, such as opening the auditorium doors at the last moment to let everyone enter at once. During days 15 until 18, Hoch spent much of her time visiting activity areas, including horseback riding, swimming, photography, canoeing, tennis, water skiing, sailing and drama. Across this spectrum, she saw great variety in a few basic processes. Activities were being shaped into regular, expected patterns by the interplay of past experience and new approaches. Novice counselors – and very occasionally novice campers – reasoned about the requirements of the activity and drew on experiences elsewhere. Returning campers and counselors mostly steered the action toward what was remembered from earlier years, though occasionally they introduced novelties by critiquing prior practices. In many of the venues, as in the conflicts in setting up dining room service tables, there were recurring tensions over replicating prior Poplar Grove experience or following judgments of new people with experience from other domains. One at the waterfront centered on whether lifeguards

142

Organizational routines and stability in organizations

should remain standing (the interpretation of Camp practice by the assistant director) or could sit (the past experience of the counselors at prior worksites). At the horse corral, it was whether campers in each activity period should saddle and unsaddle horses, or whether that should be done only at the beginning and end of the day. Observation day 21 was devoted to a special event, the Medley Marathon, which divided the Camp into red and blue teams for a series of competitive events. Many aspects of the day were traditions carried forward in general form from previous years. For example, there had to be mascots, flags and chants, although there was discretion in how the groups implemented these requirements. Racing with ping-pong balls on a spoon was a requisite event. (But this time wind was adjusted for by substituting golf balls.) Other standard events had to be recreated as well, such as a water-pistol candle shoot-out, and a game called ‘crab football’. The 22nd observation day, 4 July, was Hoch’s last, and was the closing day of the first Camp session. After regular morning activities, special events took over, along with cleaning and packing to depart. As usual in recent years, the play developed by the drama class was performed as part of the closing celebrations.

3.

EXPLAINING REGENERATION

Organizational Character To convey the sort of non-identical sameness that actually is achieved, we reintroduce the term organizational character.3 We use the term to denote the coherent content of the ensemble of dispositions that generates the distinctive actions of an organization. We will argue that this ensemble of dispositions resides in the individual procedural memories of organizational participants, and is coherent, persistent and regenerative. Once the concept is developed, we will relate it to existing literature on organizational identity and organizational culture. Our sense that organizations can be understood to have a property analogous to individual character revives a view that was advanced almost four decades ago by Philip Selznick (1957), but which has received scant attention since then. He argued that social processes in organizational work generate value commitments that shape organizational character, which is: the product of self-preserving eﬀorts to deal with inner impulses and external demands. In both personality and institutions, ‘self-preservation’ means more

The regeneration of organizational character at Camp Poplar Grove

143

than bare organic or material survival. Self-preservation has to do with the maintenance of basic identity, with the integrity of a personal or institutional ‘self’. (pp. 141–2)

At both the individual and organizational levels, a concept of character focuses our attention on how the actions generated by a coherent set of dispositions (or ‘impulses’) are recognized as ‘the same’ by those who know the person or organization. At the individual level, we speak of our acquaintances as having acted ‘in character’ when they react in familiar ways to unfolding circumstances. We find them to be ‘out of character’ when their reactions are unexpected in circumstances we judge to be ‘the same’. We label as ‘characteristic’ those specific actions of a person that most powerfully suggest the distinctive broader patterns of their behavior.4 Indeed, research reveals that because action patterns occur in distinctive correlated clusters we can gain enormously in our power to predict others’ behavior from glances as short as a few seconds. Ambady and her colleagues, for example, have shown that ‘thin slices’ of an individual’s behavior, presented in short video clips ranging from a minute down to as little as one second, are suﬃcient to allow observers to make sharply improved predictions of the actions and relationships of observed others (Ambady et al., 2000). The critical point here is that it is possible to apprehend individual character based on limited exposure and despite the flux of circumstances because the actions of individuals reveal considerable coherence.5 Even if successive situations are not identical, indeed even when there may be substantial novelty that calls for actions we have never seen, we have a sense of a person – and, by the extension we propose, of an organization – as a coherent bundle of action dispositions. We can then utilize this sense of coherence in predicting likely future behavior. So in Ambady’s observations on individuals, for example, predictive power comes less from particular behaviors than from ‘molar’ actions. Counting smiles will predict less well than counting, say, expressions of confidence (Ambady et al., 2001). There is, of course, no presumption that perceptions of character are always correct, only that we can understand enough of the correlation structure of experience for such perceptions to be possible and, on average, quite useful. Procedural Memory and Action Dispositions We use action dispositions as a general term to capture notions that writers in several traditions might discuss as traits, habits or skills. We regard an individual as developing over time a coherent ensemble of dispositions to

144

Organizational routines and stability in organizations

act in certain ways in certain situations – a persistent collection of premises, response tendencies, and structural capabilities that produce action with recognizable character. Because these action patterns are typically acquired and enacted without high levels of self-conscious analysis, and because each is to some degree co-adapted to other dispositions already developed, they normally form a quite durable and interdependent system. The developing psychological literature indicates that action dispositions such as habits and skills are retained in individuals as procedural memory, a form that involves low conscious awareness or articulation, long retention, and distinctive mechanisms of activation and generalization (Squire and Kandel, 1999). There is also evidence to suggest that individuals’ roles in organizational routines are stored in the procedural memory of the participants. Cohen and Bacdayan (1996) showed that as dyads gained experience in a recurring, joint problem-solving task they developed collective routines and stored their individual roles in them as procedural memories. The procedural memory of participating actors thus contributes some of its properties to collective action dispositions such as routines, practices or customs. This is the basis of our view that the concept of organizational character is more than a mere analogy to individual character, but is a genuine property of the organization. An important implication of storage in procedural memory is that one need not be consciously aware of assumptions about partners or settings as action is taken. These assumptions are stored ‘automatically’ as action occurs and prove evocative the next time a similar partner or setting is in view, providing a capacity for generalization that allows an individual to reproduce a lot of what is ‘in character’ for the organization even if the partner or setting are substantially diﬀerent in detail from previous years or encounters. A second implication of this focus on individual procedural memory is a key point of departure for our work from that of Selznick. Where Selznick was focused largely on the character-defining impact of critical decisions made by leaders at the tops of organizational charts, we are more interested in the reproduction of character through individual participation at all levels of the hierarchy. The camp counselors we observed were constantly confronted with uncertainty about how to behave in novel situations. We believe their actions, and the memories created by those actions, contributed to the overall coherence of the camp’s character. In other words, actions not identified as ‘out of character’ or explicitly corrected served as models for future behavior by the acting individual and any others who may have observed the behavior. It is the (only partially) coherent content of dispositions resulting from these mutually constraining actions and memories that gradually becomes recognizable as organizational character.

The regeneration of organizational character at Camp Poplar Grove

145

Mutual Adaptation and Ecologies of Action Dispositions A further source of expected coherence in the character of an organization is the mutual adaptation within the ensemble of action dispositions that, at the organizational level, we describe as an ecology. The term lets us indicate that recurring actions – or, as we’ll sometimes say, ‘practices’ – within the organization are interdependent and that action dispositions must develop into a reasonably eﬀective ensemble. In other words, one critical aspect of the process of becoming organized is a series of negotiations, both implicit and explicit, that are occasioned by actions. Individuals act based on multiple plausible action dispositions, choosing an approach based on what they feel and believe is the appropriate thing to do in the given situation. As is indicated above, this is often based on procedural memory of prior experience in situations felt to be similar on some critical dimension, either within or outside of the organization in question. As these actions are carried out, action dispositions that interfere significantly with others or that are unintelligible when combined with other actions will tend to be filtered out in the coevolution of the system of action dispositions, just as species that too often evoke negative feedback (being eaten or poisoned) will disappear from an ecology. This was the case in the observations above in the case where the counselor assigned his campers ‘kitchen duty’, to the frustration of Cathy, the head cook. In this negotiation process, experience accumulates through both action and narrative exchange. What remains is the set of action dispositions that are reasonably eﬀective and meaningful in the presence of other dispositions that meet the same criteria. This interdependent set then forms a coherent system, generating the organizational actions that are recognizable as ‘in character’. At Poplar Grove, the regeneration of that character is seeded to a significant degree by the minority of returning veterans. Acting on their retained dispositions, they shape the experiences of newcomers, thereby instilling similar – though not identical – dispositions, and hence regenerating the organization. This formulation raises several interesting and diﬃcult questions that are faced both by organizational participants and by organizational researchers. First is the problem of identifying dispositions. Even though dispositions are generally not directly observable, it is still possible to isolate specific actions taken within organizations and extrapolate likely dispositions from these. The inference processes are somewhat error prone, of course, but the ability of participants and outside observers to do this relies on the very remarkable powers of extracting the correlation structure that let us infer future actions from ‘thin slices’. Second, it is diﬃcult, both for participants and outside observers, to articulate precisely what

146

Organizational routines and stability in organizations

constitutes a ‘reasonably eﬀective and meaningful’ action disposition. One possible criterion is that a ‘reasonably eﬀective’ action disposition must not interfere significantly with others, though even ‘significant’ can be said to be highly dependent on the organizational context. In a highly interdependent and minutely controlled organization such as a semiconductor ‘clean room’, the notion of a significant failure looks quite diﬀerent from how it might look in a more fault-tolerant setting such as Camp Poplar Grove. This also raises the issue that diﬀerent organizations will have diﬀerent levels of tolerance for deviation from ‘expected’ behavior. ‘Reasonably eﬀective’ action dispositions are likely to be diﬀerent in an organization comprised almost entirely of newcomers than in one comprised almost entirely of veterans. This view departs from classical formulations of adaptive or evolutionary processes in which fitness is attributed to the direct eﬀect of actions on an environment external to the organization (many examples are provided in Aldrich, 1999). In our case, action patterns (or, more precisely, their underlying dispositions) are shaped mostly by their relation to other action patterns, while the resulting ensemble of dispositions is required to function well enough in the external environment to generate resources needed for the system’s survival or growth (Axelrod and Cohen, 1999). Character in Relation to Other Perspectives Of course, a new terminology does not erase the problem of sameness or identity, which is, after all, one of the deepest, most enduring of philosophical issues. That the world still cites both Heraclitus and Ecclesiastes is testimony to that. While the problem is not solved by locating it in a context of learning, remembering, perceiving, recognizing and enacting of ecologies of dispositions, it is usefully transformed. We can’t say very precisely how coherence suﬃcient for recognition is achieved or discerned, but these are fundamental questions which contemporary psychology seeks to answer. Even as we wait for more detailed explanations of the underlying individual and group processes, however, we can align organizational theories with the clear fact that this coherence is achieved. Useful insight is possible with what we know now and can readily observe. There is considerable overlap of the ideas we gather under the label ‘character’, and several themes are discussed in the literature of ‘organizational identity’ and ‘organizational culture’ that bear on sustaining organizational action patterns (Druckman et al., 1997; Fiol, 2002; Martin, 1992; Miller and Jablin, 1991; Trice and Beyer, 1993; Weick and Gilfallan, 1971; Whetten and Mackey, 2002). A detailed treatment of our perspective in relation to this work can be found in Birnholtz et al. (2007).

The regeneration of organizational character at Camp Poplar Grove

147

The terminology we’ve developed, however, can now be used to state our perspective: a coherent system of mutually adapted action dispositions forms an ecology that has the property of organizational character. The question intriguing us can now be reformulated: how does an ecology of dispositions – and consequent practices – reproduce itself and maintain its recognizable organizational character? This is the topic we take up in the next section.

4.

SOME REGENERATIVE PROCESSES

We can now abstract some features of Poplar Grove’s regeneration that may generalize beyond the one annual cycle that was observed. Specifically, we have observed four such processes, which will be illustrated below. In varying degrees, these processes contribute to the regeneration of coherent systems of action dispositions by serving two primary functions. First, they allow for the more or less direct transfer of knowledge, experience and practice from more experienced to less experienced staﬀ. Second, they allow for the filtering of action dispositions to arrive at a set that is reasonably eﬀective, meaningful and coherent with other dispositions. The Primacy of Demonstration One critical aspect of regeneration is the transfer of skill and experience from the more experienced core of the community to newcomers, such that they will be able to do their jobs in a way that is both satisfactory to the generic requirements of the activity area, and considered to be acceptably ‘in character’. The severe diﬃculty of regenerating so many practices with so few experienced members is endemic to seasonal organizations, as evidenced in the strong focus on core administrative personnel in the camp opening process that we find in articles and manuals on camp administration (Ball and Ball, 2004; Dimock, 1948; Drought, 1943; Gibson, 1923; Goodrich, 1959; Hamilton, 1930; Leiken, 2000; Leiken and Riggio, 2002). In Hoch’s observations, demonstration emerged as an essential aspect of the transfer of skills and practices to newcomers. In some ways, this is not surprising and confirms existing findings suggesting that new members of an organization learn their jobs by watching and learning from their peers and supervisors (Feldman, 1981; Lave and Wenger, 1991; Miller and Jablin, 1991). First, demonstrations were repetitive and increasingly representative of full-scale camp life. As is illustrated above, the staﬀ during training week

148

Organizational routines and stability in organizations

gradually adopted a schedule and set of routines that were increasingly similar to those to be used once the campers arrived. One of the first tasks during training week, for example, was setting up the area called ‘flagpole’, after which staﬀ began gathering there before meals and for meetings. Whether or not it was their conscious intent, this got the staﬀ in the habit of gathering at flagpole and demonstrated an important practice of everyday camp life. This experience, and action disposition to ‘meet at flagpole’, could then be drawn on in later moments of uncertainty. In moments of uncertainty about where to go for a new event or activity of uncertain location, most people at camp would default to flagpole. Even if the activity turned out not to be there, the probability of running into somebody there who knew where to go was high. Demonstrating this core behavior early in staﬀ training allowed this coherence to persist. Moreover, the continued use of the word ‘flagpole’ to describe this core behavior helps regenerate the Camp’s characteristic combination of flexibility with respect for the past. Second, demonstrations were multi-staged. They first involved experienced members of the Poplar Grove community, then new staﬀ, and finally the campers. In an anecdotal account, one seasoned camp director has referred to this aspect of training as ‘planting seeds’ with his staﬀ (Jacobs, 2002). Indeed, as newcomers to a strange organization, new staﬀ are hungry for any and all information they can use in understanding how they will be expected to act in this new setting. At Poplar Grove, this was evident in the observations above in the curiosity of the new staﬀ, who very frequently wondered if they were doing things the ‘right’ way. Small, early demonstrations and discussions of appropriate behavior had a strong impact that lasted throughout the summer. A third property of demonstration was that it relied on the memories of experienced staﬀ and campers. To illustrate, consider Brian’s eﬀorts to reimplement past practices in the ‘carnival night’ described above. The past practices ‘stuck’ in that several planners of subsequent auditorium-based activities (for example, dances, plays, casino nights) that summer began their events with a last-minute ‘grand entry’. Thus, we see that the demonstration of a known and eﬀective disposition to begin auditorium activities in a particular way appeared to have an impact on how subsequent events unfolded under similar, but not identical, conditions. Cascading of Guidance Another key regeneration process was the ‘guidance cascade’. These cascades were extremely common, particularly early in Hoch’s observations. Knowledge of procedures was distributed among the returning administrators and other staﬀ, and was often regenerated in new staﬀ not by direct

The regeneration of organizational character at Camp Poplar Grove

149

communication, but by advice from other participants who had heard, or surmised, the answer to an arising question. It was simply not possible for the experienced group to convey the full detail of their tacit knowledge to those in new roles. We see examples of significant cascades very early on in Hoch’s observations, for instance as she was helping set up the kitchen. Rick’s advice to her was interesting in that it was sometimes very specific (for example, ‘you can put the glasses on that shelf’), but occasionally ambivalent (for example, ‘just put it anywhere because when she [Cathy] comes in she will just put stuﬀ the way she likes it’). This episode highlights an important diﬀerence between a cascade and traditional hierarchic flow of information. In a hierarchy, Hoch would be receiving advice only from Cathy, who would be able to provide definitive guidance. In a cascade, however, limited numbers of experienced staﬀ and staggered arrival times (for examle, of Cathy) meant that advice frequently came not from an authority figure, but rather from somebody who was relying on memory to provide guidance that seemed reasonable. The sort of transfer that occurs via a cascade is neither a detailed set of instructions about how exactly one is to perform one’s job nor a reliable indicator of acceptable behavior. Rather, it is a higher-level description of how things once looked or were done, or how they might be done now. An important concomitant of this observation is what we described above as the ‘leadership of the one-eyed’, as with Lisa’s ability to answer staﬀ questions. What is particularly interesting is that very small amounts of additional experience translated into significant influence. Even Lisa, a new staﬀ member who had been at Camp only a few extra days, served as a source of valuable information about skills, practices and what may or may not be ‘in character’. A second concomitant of information cascades is action in the face of conflicting sources of guidance. Hoch often found herself in this situation and was uncertain about how to proceed, as when she received contrasting advice from Jeremy and Rick about where to put supplies in the kitchen. These situations were often resolved by trying to estimate ‘who should know best’ by taking into account knowledge about the prior experience and status of the sources. These episodes are also more common and significant in the seasonal regeneration situation which does not meet the usual conditions of stable hierarchy. There the conflicts could be resolved by asking ‘who’s my boss?’. When the advisers are co-workers without direct authority, however, the newcomer has a much more open problem, which increases reliance on the actor’s own experience and dispositions. If advisors don’t agree, then it may be an occasion to explicitly apply prior experience, or to do what ‘feels right’.

150

Organizational routines and stability in organizations

Such frequent moments of indeterminacy, oﬀering as they do the opportunity to bring new patterns of action into play, would seem to make it likely that organizational character will dissipate over time. But in our observations, actors with conflicting sources of advice made a serious eﬀort to intuit what solution made sense in the Poplar Grove context. They didn’t always succeed at selecting the ‘right’ action, of course, but their loyalty to recreating Poplar Grove strongly channeled the regeneration process. They were, in eﬀect, implicitly striving to act ‘in character’, and, by doing so, they were reproducing it. Bursty Communication A third property of regenerative processes is the nature of communication that we saw at Poplar Grove. The volume and frequency of communication were significantly constrained by the size of the camp, with some activity areas a 10 minute walk (twice that for certain 7-year-olds) from each other, and by the need to supervise the campers constantly once they had arrived. The Camp provided little in the way of communications infrastructure, save for hand-held radios for core administrative staﬀ and cellular telephones in high-risk areas. All of this meant that virtually all communication occurred in relatively brief, face-to-face encounters that we refer to as ‘bursts’. Bursts provided newcomers with ‘just enough’ information to act in ways that were unlikely to result in conflict with other actions, and to allow for the regeneration of character by seeding a framework of ‘sameness’, while allowing for slight, inevitable variations in interpretation. We observed two types of communication bursts: collective and individual. Collective bursts Collective bursts occurred when ‘everyone’ was gathered, usually either at ‘flagpole’ before each meal or at weekly staﬀ meetings. In our terms, the goal of collective bursts was to reduce the probability of action disposition conflict by providing everyone with identical information and shaping a set of shared dispositions. Collective bursts were characterized by their rapid nature, broadly applicable contents, and regular occurrence. By ‘rapid’ we mean that collective bursts generally did not take long, and often addressed a large number of topics, such that the outcome was a rapid stream of short descriptions pertaining to many topics. Flagpole, for example, was the one place where the entire Camp gathered on a regular basis, and was therefore the only reliable way to spread verbal information to the entire community. What occurred was the rapid communication of just enough information – one hopes – to provide the requisite framework for seeding the desired outcome.

The regeneration of organizational character at Camp Poplar Grove

151

By ‘broadly applicable’, we mean that collective bursts were relevant to a large number of people, which often meant that the listeners had to extrapolate the details for their specific area of Camp. For example, an administrator might announce that ‘safety needs to be an ongoing concern for all staﬀ’ rather than saying specifically to ‘watch for open-toed shoes at the horse corral and campers not wearing life jackets in their kayaks at the beach’. The implication is that the collective burst is about using generic information to set up the necessary conditions for people to make desired extrapolations for their specific areas. This, of course, also assumes some basic level of knowledge on the part of the staﬀ. A generic announcement about safety would not be likely to have an eﬀect on the behavior of a counselor unaware that campers were required to wear life jackets while kayaking. Finally, by ‘regular occurrence’ we mean that collective bursts tended to be recurring. Flagpole, one key opportunity for collective bursts, occurred before every meal at regularly scheduled times. People distributed all over Camp counted on being together at these times, and on the opportunity to get needed information. Individual bursts Unlike collective bursts, individual bursts were targeted at individuals or small groups. Individual bursts are unique among the regenerative properties that we have discussed so far in that they were one of the few means by which feedback and correction could be provided once action occurred. Rather than trying to rapidly or generically convey ‘just enough’ information to prospectively seed a desired framework, individual bursts were characterized by the conveyance of a detailed kernel of information or experience from the speaker to the listener(s). One example of this above was quick advice from Kelly about how to handle activities on the rainy day. In that instance, Kelly was not only telling the new staﬀ members what to do, but also conveying some critical aspects of her past experience with rainy days. Her clear implication was that it was most important to know where the campers were, and conducting the activity itself may be secondary. We characterize these interactions as bursts because they are diﬀerent in important ways from a normal dialog or feedback cycle. In some ways, though, these diﬀerences are more pronounced in the way the information is interpreted than in the content. In other words, the lack of an observable and functional instantiation of the organization renders individual bursts more important to new members than similar conversations might be in a more conventional and continuous organization. Individual bursts were unlike collective bursts in that they did not occur regularly, but as

152

Organizational routines and stability in organizations

opportunities arose. As a result, they tended to apply to specific situations, rather than general classes of activities. Applying Generic Skills in Context Another critical process that we observed was individuals mapping their generic skills and experience from a range of settings onto specific situations at Poplar Grove. Poplar Grove staﬀ are recruited, in part, based on their experience and ability in child care and various program activities. Among others, state law mandates qualified lifeguards at the waterfront, experienced riders at the horse corral, and people with commercial cooking experience in the kitchen. These staﬀ must then determine how the familiar components of their job are to be done at Poplar Grove. In other words, they must combine their existing knowledge and experience in a particular domain (their ‘domain knowledge’) with what they are learning (via the above processes) about how things are done at Poplar Grove (their ‘organization knowledge’). As was illustrated above, we found that job ambiguities at Poplar Grove were typically resolved by asking more experienced staﬀ for more information, or improvising a plausible solution. Asking a more experienced colleague or supervisor for more information is a common strategy for workplace newcomers, and was common in Hoch’s observations (Miller and Jablin, 1991). Occasionally, however, even asking for explicit directions raised ambiguity, as when Hoch received diﬀerent answers from diﬀerent people about how to arrange supplies in the kitchen. A second strategy that we observed was the trial of the plausible. Here, people confronted with ambiguity drew on some combination of their knowledge of Poplar Grove (or even the more general class of ‘American summer camps’), their prior experience in a particular domain (for example, lifeguarding) and information from the immediate environment to improvise a plausible solution to a problem. In some cases, as we saw above, such improvisations ‘stuck’ – perhaps because they were simply not noticed, thought by observers not to matter, or were perceived as improvements on previous practices. In some ways, this selective improvisation process shows how organizations can change while retaining the coherence we refer to as character. While the first three regeneration processes above center on conveying information that will achieve ‘sameness’ in action, this last process oﬀers significant opportunities for change. Our observations suggest two factors that are important in understanding how improvisations take hold. The first is the domain experience of the improviser. Where a newcomer knows a great deal about a domain, perhaps more than any other member, he or she may be more likely to dominate a disposition that has historical

The regeneration of organizational character at Camp Poplar Grove

153

roots in the organization. This, of course, also depends on the improviser’s persistence, the organization’s willingness to adopt novel practices, and on the next two properties. Domain experience played a frequent role in Hoch’s observations, particularly with Cathy, the head cook, who had substantial experience in commercial kitchens and, though she wanted to respect past camp practice, had definite ideas about how she wanted to operate the kitchen. Second is the visibility of the improvised actions. More visible actions, such as those performed in front of others or that depend on input from others, tended to be corrected more often than those performed independently. This was evident above in the discussion of whether the horses should be saddled before each period or at the start of each day. In the end, the horses were saddled each period for the remainder of the summer, because the practice conflicted only with the assistant director’s desire for the campers to have more riding time – and the corral was suﬃciently eﬀortful to observe that he did not choose to make frequent correction a priority. Such changes did not occur at the waterfront, where many practices are mandated by law due to inherent risks, and administrative observation is more frequent. Indeed, attention is also directed by experience and institutional requirements. At the waterfront, for example, the State of Michigan imposes rules (for example the ‘buddy system’ for swimming) that must be enforced by the Camp staﬀ. This is verified at least once per summer via on-site inspection. Thus, there is a strong incentive for administrators to focus on the waterfront, correct anomalous practices and ensure that the rules are being enforced.

5.

ACTION DISPOSITIONS THAT PERSIST IN A WORLD OF FLUX

The view of regeneration we have oﬀered is broadly in alignment with treatments of organizational action laid out in recent years by a number of scholars working in the traditions of structuration, practice and activity theories (Adler, 2003; Blackler, 1995). These are approaches that see structure as being recreated through the action taking place within it, and that emphasize human agency in context, and therefore the possibilities for change (Feldman, 2000; Feldman and Pentland, 2003; Hargadon and Fanelli, 2002; Orlikowski, 2002; Tsoukas and Chia, 2002). Our work brings us to many points of agreement with their accounts, but also discloses some diﬀerences of perspective that, we think, avoid some diﬃculties encountered in the work done so far.

154

Organizational routines and stability in organizations

Careful observation convinces all these researchers that action in organizations can be – and most commonly is – recognizably patterned without being static or mindless. These observations raise what we have called the paradox of the (n)ever-changing world: how can an organizational action pattern be ‘the same’ and still be appropriate and eﬀective in a world that is never the same? Organizations don’t usually engage in established practices or enact routines with no attention to the purposes of those doing the work or with no thought about the specific circumstances of action. This would be the analog of a driver setting out without a destination and keeping his eyes closed. And yet. . ., we do sometimes drive home when we intended to stop at the store. And a very common complaint of organizational ‘change agents’ is that established practices somehow will not adjust to the legitimate but novel requirements of context, customers or supervisors. Graham Allison’s (1971) classic study of the Cuban missile crisis reports a striking example in which Russian soldiers were told by oﬃcers on their ships to disguise themselves as civilians so as to avoid being counted by spies on the Havana docks. They did so, and then, after going down the gangway, formed into ranks and marched away. All these accounts attack the problem by devising language for rendering the interplay of what we know with how we act. Orlikowski argues that we should distinguish knowledge from knowing, which ‘is constituted in everyday practices’ (2002: 251). Feldman and Pentland (2003) distinguish the ostensive aspect of routine, the explicit concept of routine held by organizational members, from the performative aspect, the actions that members undertake in executing the routine in a real context. Hargadon and Fanelli (2002) distinguish latent knowledge (scripts, goals and identities) from empirical knowledge (artifacts, tools and routines). If one can generalize about how our account diﬀers from these, it is in our emphasis on the procedural memory basis of habitual or skilled action in humans. We rely heavily on the distinction of procedural from declarative memory now taking shape in psychological research (Anderson and Fincham, 1994; Squire and Kandel, 1999). We take habitual dispositions to be fundamental to action, and thoughtful analysis to be a vitally important mechanism of correction and – often enough – of improvement. This leads our attention to the mechanisms of transfer of skill discussed in our fourth section. In this light we can highlight some diﬀerences of detail with the studies we have mentioned. We believe, unlike Feldman and Pentland, that there can be recurring and recognizable organizational action patterns without what they call ostensive elements. Indeed, patterns that develop in this way

The regeneration of organizational character at Camp Poplar Grove

155

seem quite normal. There may be some activities in the Camp that meet the definition they give of routine, with well-developed ostensive and performative aspects, but much Camp activity seems to regenerate without any significant agreement on names for, or shared definitions of, an ostensive aspect of large action patterns. Our emphasis is diﬀerent from Orlikowski’s (2002) in that she identifies practices as rather general functional categories that interviewees express to her, such as ‘aligning eﬀort’ and ‘sharing identity’. In the case we have studied it appears more natural to take practices as our campers and counselors do, as setting the service table for lunch, saddling the horses, or setting up the swimming dock. We see organizational character as the dispositions that shape these observable practices and are reshaped by them, but not as identical to them. Orlikowski does report that at the firm she has studied there is a notion of doing things ‘the Kappa way’. We are certainly agreed that such beliefs have tremendous significance, and we try to capture them in our discussion of organizational character, but we think it may be confusing to define ‘the Kappa way’ as a behavioral practice rather than as a coherent ensemble of action dispositions. Tsoukos and Chia (2002) are quite similar to our approach in their emphasis on underlying habit and disposition. In their account ‘actors are conceived as webs of beliefs and habits of action that keep reweaving (and thus altering) as they try to coherently accommodate new experiences’. But they give such overwhelming priority to change in their argument that they end up using ‘organization’ interchangeably with stasis. They are heavily on the Heraclitus side. Indeed they quote him. But they don’t quote Ecclesiastes in response. Our diﬀerence with their approach is that by marshalling the concept of character we can provide an account of continuity despite flux. Our emphasis on the habitual dispositions that are the basis of most individual and (hence) organizational action can be captured by an apothegm of John Dewey’s (1922), remembering that by ‘instinct’ he meant what we would label ‘emotion’: ‘Man is a creature of habit; not of reason, nor yet of instinct.’ And, although we have touched on them only in passing, leaving a fuller account of the contributions of emotions to organizational regeneration for a later day, their interplay with the ecology of practices is every bit as subtle and consequential as the interplay of habit and reason. Though each perspective described above casts the issue in a distinctive form, all of these papers wrestle with some variant of the problem of reconciling the authors’ field observations of agency and change with the mindless and static connotations of our language for ‘routinized’ action. And each of them works to enrich our conceptions of the connection between what organizational actors know and the patterned actions they

156

Organizational routines and stability in organizations

undertake. It is profoundly diﬃcult conceptual territory, but a better map of it will open new possibilities for organizational research and design, and so it merits our overlapping inquiries. The unusual angle of our fascination with regeneration and organizational character has led us to what may be some fresh ideas. We hope they will stimulate new rounds of discussion.

NOTES 1. Portions of this chapter appeared previously in Birnholtz et al. (2007). 2. International students very commonly serve as staﬀ member in US summer camps. They bring even less prior knowledge to the regeneration problem than their first-time-ascounselor US counterparts. 3. There are many formal definitions of character, most of them turning on the idea of typical traits. A revealing definition in the Oxford English Dictionary is ‘the individuality impressed by nature and habit on man or nation’. This one nicely applies ‘character’ to both persons and collectives, and ties the concept, as we do, to the power of underlying regularities of action (that is, habitual dispositions) to form distinctive and coherent entities (individuals). 4. We use ‘organizational character’ in a sense quite diﬀerent from that of Bridges (2000) which clusters overall organizational decision making styles in 16 categories derived from the Meyers–Briggs personality inventory for individuals. Our focus is on much more finegrained regularities of an organization’s actions in its native context. 5. Selznick (1957: 138–9) uses ‘integrity’ for this same idea.

REFERENCES Adler, P.S. (2003), ‘Practice and process: the socialization of software development’, paper presented at the Academy of Management, Seattle, WA, USA. Aldrich, H. (1999), Organizations Evolving, Thousand Oaks, CA, USA: Sage Publications. Allison, G.T. (1971), Essence of Decision: Explaining the Cuban Missile Crisis, Boston, MA, USA: Little, Brown. Ambady, N., F. Bernieri and J. Richeson (2000), ‘Toward a histology of social behavior: judgmental accuracy from thin slices of the behavioral stream’, in M. P. Zanna (ed.), Advances in Experimental Social Psychology, volume 32, New York, USA: Academic Press, pp. 201–72. Ambady, N., D. LaPlante and E. Johnson (2001), ‘Thin slice judgments as a measure of interpersonal sensitivity’, in J. Hall and F. Bernieri (eds), Interpersonal Sensitivity: Measurement and Applications, Mahwah, NJ, USA: Erlbaum, pp. 89–101. Anderson, J.R. and J.M. Fincham (1994), ‘Acquisition of procedural skills from examples’, Journal of Experimental Psychology: Learning, Memory and Cognition, 20, 1322–40. Axelrod, R.M. and M.D. Cohen (1999), Harnessing Complexity: Organizational Implications of a Scientific Frontier, New York, USA: Free Press.

The regeneration of organizational character at Camp Poplar Grove

157

Ball, A. and B. Ball (2004), Basic Camp Management, Martinsville, IN, USA: American Camping Association. Birnholtz, J., M.D. Cohen and S.V. Hoch (2007), ‘Organizational character: on the regeneration of Camp Poplar Grove’, Organization Science, 18(2), March–April, 315–32. Blackler, F. (1995), ‘Knowledge, knowledge work and organizations: an overview and interpretation’, Organization Studies, 16(6), 1021–46. Bridges, W. (2000), The Character of Organizations, Palo Alto, CA, USA: DaviesBlack Publishing. Cohen, M.D. (2007), ‘Reading Dewey: reflections on the study of routine’, Organization Studies, 28, 773–86. Cohen, M.D. (1999), ‘Commentary on the Organization Science Special Issue on Complexity’, Organization Science, 10(3), 373–6. Cohen, M.D. and P. Bacdayan (1996), ‘Organizational routines are stored as procedural memory: evidence from a laboratory study’, in M.D. Cohen and L. Sproull (eds), Organizational Learning, Thousand Oaks, CA, USA: Sage Publications, pp. 403–29. Cohen, M.D., R. Burkhart, G. Dosi, L. Egidi, M. Marengo, M. Warglien et al. (1996), ‘Contemporary issues in research on routines and other recurring action patterns of organizations’, Industrial and Corporate Change, 5(3), 653–98. Dewey, J. (1922), Human Nature and Conduct: An Introduction to Social Psychology, New York, USA: H. Holt and Co. Dimock, H.S. (1948), Administration of the Modern Camp, New York, USA: Association Press. Drought, R.A. (1943), A Camping Manual, New York, USA: A.S. Barnes & Co. Druckman, D., J. Singer, H. Van Cott, J. Singer and H. Van Cott (1997), Enhancing Organizational Performance, Committee on Techniques for the Enhancement of Human Performance, National Research Council, Washington, DC, USA: National Academies Press. Eisner, M.D. (2005), Camp, New York, USA: Warner Books. Emerson, R.M., R.I. Fretz and L.L. Shaw (1995), Writing Ethnographic Field Notes, Chicago IL, USA: University of Chicago Press. Feldman, D.C. (1981), ‘The multiple socialization of organization members’, Academy of Management Review, 6(2), 309–18. Feldman, M.S. (2000), ‘Organizational routines as a source of continuous change’, Organization Science, 11(6), 611–29. Feldman, M.S. and B.T. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48(1), 94–121. Fiol, C.M. (2002), ‘Capitalizing on paradox: the role of language in transforming organizational identities’, Organization Science, 13(6), 653–66. Gibson, H.W. (1923), Camp Management: A Manual for Camp Directors, Cambridge, MA, USA: Murray Printing Company. Goodrich, L. (1959), Decentralized Camping: A Handbook, New York, USA: Association Press. Hamilton, A.E. (1930), Boyways: Leaves from a Camp Director’s Diary, New York, USA: The John Day Company. Hargadon, A. and A. Fanelli (2002), ‘Action and possibility: reconciling dual perspectives of knowledge in organizations’, Organization Science, 13(3), 290–302.

158

Organizational routines and stability in organizations

Jacobs, J. (2002), ‘Planting seeds with your staﬀ: growing your camp’s culture’, Camping Magazine, 75, 20–23. Lave, J. and E. Wenger (1991), Situated Learning: Legitimate Peripheral Participation, Cambridge, UK: Cambridge University Press. Leiken, J. (2000), ‘Everything you say and do matters: teaching staﬀ to put campers first’, Camping Magazine, 73, 29–32. Leiken, J. and J. Riggio (2002), ‘Creating your ideal camp culture’, Camping Magazine, 75, 16–19. Martin, J. (1992), Cultures in Organizations: Three Perspectives, New York, USA: Oxford University Press. Miller, V.D. and F.M. Jablin (1991), ‘Information seeking during organizational entry: influences, tactics, and a model of the process’, Academy of Management Review, 16(1), 92–120. Orlikowski, W. (2002), ‘Knowing in practice: enacting a collective capability in distributed organizing’, Organization Science, 13(3), 249–73. Pentland, B.T. and M.S. Feldman (2002), ‘Organizational routines as a unit of analysis’, paper presented at the conference Empirical Research on Routines in Business and Economics: Towards a Research Program, Odense, Denmark. Selznick, R. (1957), Leadership in Administration, New York, USA: Harper and Row. Squire, L.R. and E.R. Kandel (1999), Memory: From Mind to Molecules, New York, USA: W.H. Freeman & Co. Trice, H.M. and J.M. Beyer (1993), The Cultures of Work Organizations, Englewood Cliﬀs, NJ, USA: Prentice-Hall. Tsoukas, H. and R. Chia (2002), ‘On organizational becoming: rethinking organizational change’, Organization Science, 13(5), 567–82. Weick, K. (1995), Sensemaking in Organizations, Thousand Oaks, CA, USA: Sage Publications. Weick, K. and D.P. Gilfallan (1971), ‘The fate of arbitrary traditions in a laboratory microculture’, Journal of Personality and Social Psychology, 17(2), 179–91. Whetten, D.A. and A. Mackey (2002), ‘A social actor conception of organizational identity and its implications for the study of organizational reputation’, Business and Society, 41(4), 393–414.

8.

Uncovering inertia: ambiguity between formal rules and routines of interaction Martijn van der Steen

INTRODUCTION Institutional change arises from interactions between rules and routines (Burns and Scapens, 2000). However, both institutions1 and routines are considered to display a degree of resistance to change. This resistance to change is often attributed to the presence of inertia. But so far, the notion of inertia has been used to a limited extent in studies of institutional change. Rules and routines are known to interact, but how inertia intervenes in this interaction is not explored in any great length. This chapter aims to explore the emergence of inertia between formal rules (as embedded in the management control system) and routines. In particular, it will discuss how ambiguity in the dialectic of rules and routines is one way in which inertia can emerge. Drawing on a case study of the Dutch cooperative Rabobank, we will illustrate two instances where we found ambiguity in the rules–routines dialectic. The discussion of these two instances illustrates particular ways in which inertia can arise from the ambiguity caused by changes in the rules–routines dialectic. Next, the chapter will present a theoretical section, which will first discuss rules and routines. Then, it will define routines as ‘scripts of interaction’ and explain the relevance of this focus on scripts. Finally, the section will expand on inertia and why ambiguity can give rise to inertia in the context of rules and routines. Then, the following section will introduce the Dutch Rabobank, the setting in which the two instances of inertia were located. Next, we will provide some remarks on the methodological choices made for this chapter. The part thereafter will discuss the two instances of inertia through ambiguity that were identified at the Rabobank. These two instances are illustrative of the ways in which new formal accounting rules and regulations can create ambiguity in relation to existing routines of behaviour. They also demonstrate that formal rules can create ambiguity 159

160

Organizational routines and stability in organizations

on their own account, in instances when no routines are available to organizational participants to guide them to act on these new rules. The chapter concludes with a discussion and a conclusion.

THEORETICAL BACKGROUND The interaction of management accounting rules and the behaviours that people actually display is an example of structuration as proposed by Giddens (1984). Structuration theory argues that social structure arises through the recurrent actions of knowledgeable human actors. Giddens views routines as the very fabric of social structure, which exists as a result of the continuous production and reproduction of action. However, Barley and Tolbert (1997) and Seo and Creed (2002) criticise the lack of attention to the temporal nature of structuration, as well as its inability to account for human agency in institutional change. Barley and Tolbert oﬀer an alternative view that addresses these concerns. Their approach was adapted by Burns and Scapens (2000) and was used to provide explanations of processes of management accounting change. These authors argue that institutional change emerges through the ongoing interaction of rules and routines. Rules are formal regulations and routines are the actual behavioural regularities that are aﬀected by rules, other routines and institutions. Both Barley and Tolbert (1997) and Burns and Scapens (2000) argue that institutional change arises from a dialectic between rules and routines, but neither clarify the role of inertia, which may impede institutional change. Therefore, the next sections will expand on the notions of rules and routines to facilitate the subsequent discussion which aims to locate inertia in the dialectic of rules and routines. Rules Burns and Scapens define rules as the ‘formally recognised way in which “things should be done”’ (2000: 6). Rules can be part of the existing institutional order and they are carried forward in time by both existing practices/ routines and in the structural features of the organization, which include standard operating procedures, manuals and accounting regulations. The change programme at the Rabobank, which will be described in later sections, introduced various new rules in the organization and was aimed at the emergence of new routines. It is important to distinguish clearly between rules and routines, to avoid confusion in studying their interaction. Sharp (1994) suggests that Standard Operating Procedures (SOPs) can be construed as routines that facilitate

Ambiguity between formal rules and routines of interaction

161

decision making in the face of complexity and uncertainty. In this, he makes no distinction between formal rules and behavioural routines. Cohen and Bacdayan (1994) and Burns and Scapens (2000) do make this distinction. They view routines as the informal counterpart of formal rules of behaviour. In their view, rules are the formalized statements of procedures (Burns and Scapens, 2000: 7). However, rules are not limited to written guidelines; they can also include many types of managerial instructions. Rules can be either direct action directives to individuals or groups, or they describe appropriate interpretations of particular events (Reynaud, 1996). Reynaud (2005) proposes that rules are unavoidably incomplete, and therefore unable to prescribe behaviours in any detailed sense. Therefore, a degree of improvisation is always associated to rule following. New rules often incite action, but they do not uniquely define which actions must be taken. As such, new rules do not necessarily lead to the formation of uniform behavioural regularities. Rather, formal rules are resources that individuals can draw upon to guide their behaviours in the performance of the routine, and also to account for and to explain these behaviours (Feldman and Pentland, 2003). Rules function as both the onset of the emergence of routines and as one of the cognitive and discursive representations of the routine. Put diﬀerently, formal rules can invoke the emergence of new routine behaviours, and they can describe these behaviours to a certain extent. But formal rules do not represent a uniform understanding of the routine by the entire organizational population. As such, new rules embody the potential for ambiguity, as they include many possible interpretations, which can lead to confusion relating to their execution. Routines There is a large amount of literature discussing the role of routines in organizations (Nelson and Winter, 1982; Cohen, 1996; Cohen et al., 1996; Feldman, 2000, 2003; Becker, 2004, 2005; Becker and Knudsen, 2004; Hodgson and Knudsen, 2004; Lazaric and Raybaut, 2005). Becker (2004) provides an overview of some of the most striking features of routines. He notes that routines are recurrent collective action patterns, which display some degree of path dependence and are often triggered by ‘something’, such as a cue. Becker furthermore mentions that routines can be invoked mindlessly, but they can also require some conscious eﬀort. Three broad interpretations of routines are in use (Becker, 2004, 2005). First, they can refer to behavioural regularities; as such routines are considered to be collective behavioural patterns. Secondly, routines can be conceptualized as formal rules, such as standard operating procedures. They are seen as formal guidelines, often documented in rule books and manuals (this work

162

Organizational routines and stability in organizations

refers to this category as ‘rules’). A third view of routines conceptualizes them as dispositions for particular behaviour or thought (Hodgson and Knudsen, 2004; Becker, 2005). In this view, routines are not observable actions or formal regulations, but dispositions making some behaviours and thoughts more likely than others. A disposition focuses on the cognitive processes leading to behaviour, rather than on the behaviour itself. One notable feature of routines that Becker (2004) discusses is patterns. Patterns refer to the regularities that characterize routines. The notion of regularity is present in many conceptions of routines (for example Nelson and Winter, 1982; Teece and Pisano, 1994; Cohen et al. 1996; Burns and Scapens, 2000), but it is unclear what constitutes these patterns. Becker (2004) discusses four diﬀerent types of patterns:2 action, activity, behaviour and interaction. Action and activity are often synonymous, denoting all forms of individual and collective activity. Behaviour is a subset of action, referring only to those actions that are observable and emerge as a response to some stimuli. Finally, interaction is also a subset of action; it is action that involves multiple actors. Drawing on Barley and Tolbert (1997), this chapter views interaction as the basis of routines. Routines can then be defined as: collective patterns of habitual interaction, which can be mindless or require conscious eﬀort, and which require a ‘trigger’ or ‘cue’ to be invoked. Barley and Tolbert argue that these interactions are guided by scripts, which they define as: ‘observable recurrent activities and patterns of interaction characteristic of a particular setting’ (1997: 98). An important distinction between scripts and routines is that the latter are collective phenomena, whereas the former are primarily viewed as individual phenomena (see: Schank and Abelson, 1977; Gioia and Poole, 1984). As such, a script is a useful concept for exploring the individual source of inertia underlying routine behaviours. Scripts Louis observes that in normal everyday action, individuals operate in ‘a kind of loosely preprogrammed, nonconscious way, guided by cognitive scripts’ (1980: 239). She suggests that conscious thought does not play a major part in these activities. Porac et al. note that ‘many experienced employees perform their work with highly routinised behavioural patterns and thus may not engage in much causal reasoning simply because work has become “scripted”’ (1983: 286). These writers refer to scripts, but related notions have also been used, including schema (Weick, 1979) and habitualization (Berger and Luckmann, 1979). On an individual cognitive level, scripts are at the basis of observable routine behaviours (Barley and Tolbert, 1997). Routines (as patterns of interaction) reside in the

Ambiguity between formal rules and routines of interaction

163

individual’s cognitive structure as scripts. As such, scripts can be viewed as the cognitive counterparts of routines. Scripts are cognitive knowledge structures. This means that they hold information on appropriate sequences of behaviour for particular circumstances. Scripts allow organizational participants to economize on cognitive resources by relying on scripts in much of their day-to-day behaviour. Although scripts can be invoked and performed subconsciously and they can consist of tacit knowledge, the opposite can also be true. As such, we can distinguish between two types of scripts: tacitly invoked scripts and consciously invoked scripts. The former can be regarded as cognitive automatisms that allow organizational participants to perform familiar tasks in a highly ‘automated’ fashion. These scripts are diﬃcult to access consciously, as they are primarily available as tacit knowledge and they are invoked based on tacit criteria. Examples include driving a car and performing basic manipulations on a computer. The latter, consciously invoked scripts, involve conscious selection and performance of scripts. The individual needs to choose between a series of alternatives associated to particular circumstances. For example, the ‘go’ decision for a Space shuttle launch is a consciously invoked script. We will briefly explain these two categories of scripts further, as we will make some concluding remarks on their potential association to diﬀerent levels of inertia. Tacit scripts Tacitly invoked scripts are invoked on a subconscious level (see Schank and Abelson’s (1977) restaurant script, which describes the sequence of events when dining at a restaurant). Tacit scripts occasionally lead to suboptimal decision-making. Gioia and Poole note that the recall of events for a similar or prototypic situation provides the decision maker with the script for understanding and predicting the outcome of the decision. Scripted decision making, therefore, is eﬃcient decision making, but not necessarily good decision making. The scripting of decision situations has an obvious drawback: it can induce a failure to be aware of the fine-grained diﬀerences that distinguish a current decision problem. . . . This is because the process of deciding is based on a protoscript, rather than a step-by-step accounting of the uniqueness of events relevant to the present situation. (1984: 454)

In general, this holds for schemas, which are cognitive knowledge structures encompassing scripts, plans, categories, prototypes and implicit theories (Woﬀord, 1994: 181). As scripts are part of schemas, they have many similar qualities. Harris (1994) proposes that the schema-directed nature of the perceptual process ‘lessens the frequency with which schema inconsistent information is discovered and made conscious. The very

164

Organizational routines and stability in organizations

nature of schemas act to ensure that drastic challenges to their validity seldom arise.’ (p. 311). This is because new knowledge is usually fitted in existing schemas. New knowledge is formulated in terms of old schemas, and therefore helps to sustain these schemas, as it reduces the probability that this knowledge challenges existing schemas. An example of these types of schemas can be found in the recall schema at Ford. Gioia (1992) reports that full recalls of problematic types of cars would only occur if their defects fit recognizable patterns. Any deviation in these patterns would probably not trigger appropriate behaviour. The inability of the recall coordinator and his team to invoke an appropriate script led to a faulty product not being recalled, as the problems did not fit the pattern of a ‘typical’ recall case. Although the consequences of many tacit scripts are less dramatic, they always run the risk of being invoked on the wrong occasion. Tacitly invoked scripts can suﬀer from a mismatch between the original assumptions underlying the script and current circumstances, rendering the outcomes unpredictable and often unexpected. Consciously invoked scripts Consciously invoked scripts are those scripts that involve a degree of sensemaking (Weick, 1995). Weick and Roberts (1993) use the concept of heedful interrelating to argue that actors in social systems, where error-free operations are important, understand the relations between their actions and those of others, as well as their contributions to the system. Weick and Roberts (1993: 357) used the example of an aircraft carrier flight deck to illustrate an environment where many scripts of interaction are consciously invoked, through heedful interrelations.3 Another example is Louis’ (1980) description of newcomers in an organization. Newcomers have little knowledge of cues and associated scripts, so they engage in learning behaviour. They consciously attempt to make sense of the events in their new settings, and thus learn about the types of cues that warrant the various behaviours. In these early stages, newcomers learn to use scripts that are invoked tacitly by more experienced organizational members. Here, attempts to solve ambiguity may lead to conscious attempts at sensemaking. Scripts as Individual Resources for Routines The preceding sections have extensively discussed the notion of scripts. These sections argued that scripts are cognitive repertoires that individuals can draw upon, either consciously or in a more subconscious and automatic fashion. These cognitive repertoires are important resources for the functioning of routines. Routines as patterns of interaction (Becker, 2004) emphasize both the collective and behavioural nature of routines. In this

Ambiguity between formal rules and routines of interaction

165

Routines Tacit scripts Rules Conscious scripts

Figure 8.1

Rules and scripts

definition, routines can be invoked habitually or more consciously by the individuals involved. In order to invoke and perform a routine, individuals draw on existing scripts. Nooteboom (2000) suggests that organizational routines are made up of nodes, which consist of individual behaviours and cognitions. In doing so, he proposes a linkage between routines and scripts which we draw upon. A routine, that is a pattern of interaction, can be conceptualized as a series of nodes, each of which is substituted by the actions of individuals. These actions are often a consequence of the performance of an individual script. Yet, there are many variations in the behaviours in the routine. As Feldman and Rafaeli note: ‘individual behaviours constructing a routine cannot be expected to be identical every time. But they can be expected to conform to a typified pattern’ (2002: 328). Variations in the degree of tacitness of scripts-in-use in the routine cause variations in the routine’s capacity for change. As such, the conceptualization of routines as sequences of nodes consisting of individual scripts explains how routines can change quite dramatically as a result of the conscious processing of the outcomes of earlier iterations (or nodes) in the routine (Feldman, 2000; Feldman and Pentland, 2003). Figure 8.1 depicts the relation between rules and routines on an individual level. Formal rules are related to scripts as they form resources that people can draw upon in seeking guidance for behaviour and they allow them to explain and account for their behaviours. The individual scripts are linked together through nodes that make up organizational routines, hence the dotted box around the scripts, which signifies the routine. The scripts, associated to routine behaviour, are divided into tacitly invoked scripts and consciously invoked scripts to emphasize the varying degrees of cognitive eﬀort involved in their performances. The purpose of this discussion of rules and routines is to be able to locate the notion of inertia. The next section will discuss this notion of inertia and explain why it is often manifested by ambiguity.

166

Organizational routines and stability in organizations

Inertia through Ambiguity The set of rules and routines that constitute institutions are inert for various reasons. Inertia is traditionally defined as ‘the inability to enact internal change in the face of significant external change’ (see Miller and Friesen, 1980). Gilbert (2005) proposes a division of inertia into two components: resource rigidity and routine rigidity. The former is generally associated with path-dependency and technological lock-in. It occurs as a result of the influence of external resource providers on internal managerial choice and the presence of technologies that often lead to an exclusion of other technologies, exemplified by the dominance of the VHS video format, or the internal combustion engine (see for example Dosi, 1997; Egidi and Narduzzo, 1997 and Barnes et al., 2004). Routine rigidity has the focus here. It involves the persistence and inflexibility that often typifies routines as they, through ongoing reproduction, take on structural features (Nelson and Winter, 1982; Feldman and Pentland, 2003). Gilbert (2005) notes that an explanation of this form of inertia is that the original motivation for designing the routine has been lost, while the routine is still in use and is often being invoked automatically. It has become a set of tacitly invoked scripts. The logic of the routine has become engrained in the tacit cognitions of individuals in the organization (Tripsas and Gavetti, 2000), thereby making them less accessible to conscious change eﬀorts. These tacit cognitions are encoded in the scripts that guide the individual’s interpretations and subsequent behaviours. The empirical section of this chapter will argue that this particular form of inertia is closely related to the presence of ambiguity, caused by a tension between existing scripts (as cognitive counterparts of existing routines) and newly introduced rules. To introduce the empirical enquiry into the relation between inertia, ambiguity and the rules–script dialectic, the next section introduces the Dutch Rabobank, which is the bank where the case study of this chapter was carried out.

THE RABOBANK The Rabobank organization is one of the top three banks in the Netherlands with a balance sheet total of some 556 billion euros and employing some 56 000 people in 2006. Its extensive network of branches (1214 at the end of 2006) and its direct focus on the agricultural community are its traditional strengths. Moreover, it has a very strong credit rating, reflected in its triple A status with all three major rating agencies, a continuing source of pride for the Rabobank.

Ambiguity between formal rules and routines of interaction

167

The element that distinguishes the Rabobank most from other major banks in the Netherlands today is that it is a cooperative (Sluyterman et al., 1998). Within this ownership structure, autonomous local banks are members of the cooperative. These member banks have members of their own in the form of private persons from their local environment. A number of these members are appointed to the board of each local bank. Formally, member banks are independent and have full autonomy, except for matters related to the supervisory role of the Dutch Central Bank, for which they have delegated responsibility to the supra-local organization, Rabobank Nederland. Although the bank recognized that its cooperative structure is potentially its main competitive advantage, it also realized that the absence of shareholders has led to a number of ineﬃciencies. This realization brought an increasing focus on eﬃciency and financial results within the Rabobank. Against this background, a programme called Results Oriented Management (ROM) was introduced. ROM was a programme to improve the practices of planning and control and to introduce a sense of accountability towards customers and to each other. It had a number of structural features that were consistent with those found in standard operating procedures. It consisted of a number of discrete rules of how to plan ahead and how to be in control. For most member banks, these rules were useful for their eﬀect on employees’ behaviours, including their interactions and decision-making. As such, the ROM programme consisted of a number of steps that various organizational participants had to follow, which led to the generation of new rules of behaviour. Structured as a planning and control cycle, most member banks used ROM to bring about a behavioural change and to introduce result-orientation as mode of acceptable behaviour. The formal procedures surrounding the programme are depicted in Figure 8.2. The programme formally delivered a number of plans (strategic plan, year plan, and so on) that were formal representations of the changes in attitude that the bank wanted to achieve.

METHODOLOGY The primary method of data gathering for this work has been the case study method. The case study approach focuses specifically on understanding the dynamics present within single settings (Eisenhardt, 1989). Yin (2003) argues that within single case studies, multiple levels of analysis can exist. This study was performed using a two-level analysis. A first order analysis was held in which respondents charted the most relevant issues.

168

Organizational routines and stability in organizations

Start

Recalibrate mission

Asess strenghts & weaknesses

Asess opportunities & threats

Conclusions SWOT analysis Formulate strategic course

Determine targets of organization areas

Determine targets result areas (Compasses)

Determine spearheads next 3 years Strategic plan Set out actions coming year Year plan Translate actions to segments, staff & projects

Combine & adjust plans Segment, Staff, project plan Create budget & create agreement on control Budget Exercise control

Figure 8.2

Results Oriented Management at the Rabobank

Ambiguity between formal rules and routines of interaction

169

A second order analysis was subsequently performed, focusing on theoretical interpretations of these issues and their relations by the researcher. This approach was originally suggested by Van Maanen (1979) and used by Labianca et al. (2000) and Balogun and Johnson (2004). Although the primary unit of analysis is the individual in the specific social group in which they are located, it is the interaction between this individual and the various social structures of which they are part that form the focus in this work. The interviews were held at 14 local member banks, and at the supralocal Rabobank Nederland. At 12 of these autonomous banks and at Rabobank Nederland, the number of interviews per organization was approximately three. The interviews were mostly held with the following functions at each bank: (1) the General Director; (2) a commercial manager, that is the manager of the Corporate Clients, Financial Advice or Client Advice departments, these being the primary ‘customers’ of the ROM programme; and (3) the Manager Business Administration, who often was the designated project leader. Further interviews were conducted with employees from various departments. At the two other member banks, the process of Results Oriented Management was traced from its initial implementation for a period of two years. A total of 81 interactions were held at the participating 14 banks and at Rabobank Nederland, totalling 200 hours. Almost all of these were recorded and literally transcribed, which resulted in a dataset exceeding 1400 pages and six binders with archival data. The study traced if and how new accounting principles were institutionalized at the Rabobank (Van der Steen, 2006). The current chapter is a sub-section from this study, analysing the inertial forces that were present at member banks during implementation of the ROM programme.

RULES, ROUTINES AND AMBIGUITY AT THE RABOBANK This section will discuss the introduction of new formal rules and the resulting ambiguity. Drawing on the case study at the Rabobank, we will illustrate two instances of inertia which resulted from the presence of ambiguity. We will discuss one instance of ambiguity between existing routines and new rules, and one where ambiguity resulted from the introduction of various inconsistent formal rules. The next section discusses ambiguity between existing scripted patterns of interaction and new rules; the subsequent section focuses on an instance of ambiguity resulting from the introduction of incompatible new rules.

170

Organizational routines and stability in organizations

Adverse News The routine that we discuss here is called the adverse news routine. It involved interactions that people at a number of member banks had in reaction to adverse conditions or when they were confronted with bad news. They proceeded in a similar manner on a variety of occasions. The routine encompassed various scripts, which followed this profile: 1. 2.

3. 4.

A piece of adverse information reached one or more organizational participants. This information was given meaning through experiences with earlier information. This meaning often diﬀered from the message that was conveyed originally. Rather, it resulted from interactions through which organizational participants downplayed the implications of the information. People responded to the messenger either not at all, or in a manner that reflected the limited personal threat they felt from the news. Logical inferences stemming from the information were not well understood by the parties aﬀected as they made a diﬀerent ‘mental translation’; a diﬀerent meaning was provided.

The Rabobank Zevenhuizen-Moerkapelle provided a good illustration of this routine. The management team of the bank decided to shut down its travel agency primarily because it could not achieve a banking productivity of 1.3, meaning that for each euro in cost, it needed 1.3 euros in revenues to remain sustainable on the longer term. However, recognizing that this was not realistic in the travel industry, and considering the additional traﬃc that a travel agency generates, the bank decided that the travel agency needed to be at least self-supporting. However, the accounting system produced a productivity figure between 0.7 and 0.8, which was unacceptable to the management team. The travel agency needed to be closed. But the people working there persisted in their belief that things were not that bad. Consider this critique of the General Director of the bank: This was our situation. I thought that it was clear to everyone that this was not sustainable. We needed to closely look at our costs and revenues to see if we could approach the 1.0 productivity figure. If we could do that, or we could see light at the end of the tunnel, then we would continue operations; otherwise we needed to close down the travel agency. Looking back, I must say that it was suﬃciently clear to me, but I haven’t repeated it enough to the people. I mentioned the types of choices we needed to make, but my people must have thought: ‘It will be fine, so many people are working on the issue, performing calculations to establish the sustainability of the travel agency; the bank is obtaining good financial results, it has operated nicely over the years; it will all

Ambiguity between formal rules and routines of interaction

171

be fine . . .’ That is where the source of the massive unrest lay when we decided to shut down travel operations indefinitely. With regards to communication . . . the facts have been stated to the people involved, but they placed the information in a context that I now can vividly understand, but that was not meant to.

From his position and role as a General Director, he made judgments that were extensively communicated with members of the management team, but only to some extent with the people involved. However, the mere fact that the available information was placed in a group-defined context that is diﬀerent from that of the management team is not always problematic. In the above example it was, because the meaning given by the travel agency employees was not consistent with the intent of the messsage of the management team. Similar issues arose at the Rabobank West-Zeeuws Vlaanderen, where the new General Director conveyed the message that the bank was not performing as well as everyone believed. An employee of the Client Advice department noted that this message was simply unbelievable: ‘We honestly thought we were in great shape, our oﬃce building, the presents we received, the many Belgian customers, amongst ourselves, we were convinced that we were doing marvellous.’ However, the management team used ROM to provide counter-evidence that did not fit with current conceptions of the performance of the bank. When Results Oriented Management produced the realisation that we were not doing so well on the domestic market, compared to our sister banks, we could not believe it. Cees [the General Director] confronted us many times with the performances of our sister bank in Baarle-Nassau and provided evidence from the accounting system, such as averages of services provided per year, but it took us much time to realise that our assumptions may have been inaccurate.

The adverse news routine was a mode of interaction between people that relied heavily on the shared assumption that their organization was in an optimal condition. The events they were finding themselves in (see Weick, 1995) were interpreted and given meaning under the assumption that the current performance of the organization was satisfactory. Others, often managers and directors, provided information that they find compelling proof of an opposite view. But this proof was often questioned, allowing its implications to be dismissed. Every time managers and consultants were using accounting information stemming from the ROM processes to substantiate their views on the performance of the bank, people would argue that this accounting information was either faulty or irrelevant as it contradicted their understanding of the economic position of the bank. Basically, the adverse news routine was founded on an institution, that is, a shared taken-for-granted assumption on the state of the bank, which was

172

Organizational routines and stability in organizations

a pivotal element that allowed the routine to function. People were able to develop interactions by virtue of similar understandings they had with regards to the position of the bank. In this light, the messages of managers and consultants were literally unbelievable, and this was reinforced by interactions of people outside the management teams who all shared the assumption that, whatever was going on, it was unlikely that it had to do with the performance of the bank. The adverse news routine included scripts that reduced ontological insecurity on the part of organizational participants, in particular when they felt threatened (Schein, 1985; Busco et al., 2006). As respondents at Rabobank Zevenhuizen-Moerkapelle and West-Zeeuws Vlaanderen indicated, formal signals were downplayed by a collective sensemaking process that led to a diﬀerent outcome: ‘managers are working hard to improve results, therefore it will be fine’. The output of the accounting system and the managerial interpretations called for mobilization of all staﬀ. But employees dismissed the accounting information for the larger part as they were presented with and even created evidence to the contrary, through collective sensemaking processes that were based on scripts already present. Even more so, in the face of adverse information, the scripts which were part of the routine gained salience for the organizational participants who were aﬀected by the formal changes, such as the closing of a travel agency. The situation was typified by increasing ambiguity and employees were prone to share their views on the situation with their peers. They tended to use similar scripts to interpret or make sense of new information. This seemed to confirm their interpretation of the situation. As a result, major diﬀerences in meaning arose as managers were employing diﬀerent scripts for meaning creation from the employees that were aﬀected. Figure 8.3 depicts this ambiguity between the formal messages conveyed by the managers and by the accounting system, and the scripted interaction patterns of employees that were in place. This ambiguity forced organizational participants to downplay either the new formal rules or to alter existing scripts of interaction to resolve the apparent contradiction between formal signals and their collective interaction patterns. At the member banks under study, people most often opted to downplay the importance of the signals stemming from the accounting system. At the same time, they interpreted organizational action in terms of existing scripts: ‘people are working to resolve this situation, therefore, it will be alright’. The sentiment that formal systems were producing invalid signals was reinforced by these scripted patterns of interaction. The ambiguity resulting from a misalignment between existing patterns of interaction and the signals from new accounting practices led to the reinforcing of existing patterns of behaviour

Ambiguity between formal rules and routines of interaction

173

Formal arrangements Routines Tacit scripts Rules: accounting system indicates problem

Ambiguity Conscious scripts

Figure 8.3

Ambiguity between formal rules and various scripts

at the expense of the perceived validity of the formal accounting systems. As a result, organizational participants were not prepared to consider the probability that the formal systems were valid sources of information. The Absence of a Control Routine The second instance of ambiguity discussed in this chapter is not about a routine that was present, but rather one that was absent at a number of member banks. The interest in the absence of this routine was fuelled by frequent references to respondents’ inability to confront each other with disappointing results. They indicated that they were not used to challenging colleagues by referring to their responsibilities. Various organizational participants were especially disappointed by the limited control on the agreements made in the course of the ROM programme. This was a weakness in the programme, given the adaptations made to the second version of the ROM programme management manual: ‘Many local member banks have successfully created policy plans over the last few years. The real challenge however is to realise these policy plans by which we can achieve better results. It is for this reason that this management manual emphasises Control on the execution of the policy plans’ (emphasis in original).4 Control was about confronting people with the results obtained, oﬀering support to those who needed it, and learning from those who appeared to be successful in their tasks. These were patterns of behaviour that were not embedded in day-to-day interaction. Rather, the organization was often satisfied with limited knowledge on the causes of deviations between planned and actual results. A commercial manager at the Leeuwarden bank illustrated this.

174

Organizational routines and stability in organizations

We have never talked about how we would deal with control. We were just finished; the strategic plan was ready, and so was the year plan. And now we could track the numbers, that was the feeling then. And so we went. But how we were going to do that and what we would do if deviations from planning occurred, we did not discuss that.

The reasons for the absence of a control routine, so common in the business environment, can be attributed to the almost monopolistic position many member banks held in their local communities. Their strong embedding in local communities sometimes provided local market shares exceeding 80 per cent. Moreover, customers had a close relationship with the local member banks, whose staﬀ were from the same community and whose general director often was a notable member of the community. As such, the cooperative flourished for many years, with control exercised by the local assembly of members through its board of members and a board of supervision. But, neither of these consisted of professional managers. Moreover, these boards possessed only a fraction of the information that the General Director of a bank and his management team had. This led to the situation where control of the member bank by its owners, the members, was incomplete. Organizational members were able to develop modes of interaction that focused on a friendly communication, while confronting someone with their results was seen as an undue intrusion of personal space and a denial of the valuable position that they had in the organization. The slack that member banks had, through their comfortable market position and through the inability of the control organs to exercise extensive control, resulted in unaddressed ineﬃciencies. In the course of the programme, participants indicated that they were making plans, with which they were quite pleased. But when it came to control of those plans and the check on the achievements of those plans, people were more critical. Managers and employees of member banks in West-Zeeuws Vlaanderen, West-Drenthe, Groningen, Leeuwarden and Zevenhuizen-Moerkapelle reported that the ROM process was hindered severely by the absence of a control routine. The people were just not familiar with confronting each other with the results they achieved. This was an issue on multiple levels: it aﬀected the relations between managers and employees, between employees, between managers and even between customers and the bank. Witness to this last point is the statement by a financial controller of the Groningen bank: Diﬀerences between banks become visible . . . for example when debtors default. The criteria are not always clear, but I am sure that interpersonal elements play their part. We have had such an issue with someone. And we stuck by him. I have always wondered what it was that made us decide to continue with that

Ambiguity between formal rules and routines of interaction

175

person: was it because we felt that his business was viable or didn’t we dare to confront him with the fact that we did not believe in his venture any more.

The absence of a control routine was nothing new, but ROM exposed this absence as a problem. Many organizational participants did not know how to appeal to each other’s responsibilities, which left most agreements resulting from ROM prone to being ignored. To some member banks, the absence of a routine that dealt with being accountable to one’s peers and one’s managers was a cultural trait that needed change. Therefore, a number of member banks started to implement more explicit control structures in the hope that these would lead to a behaviour that included more explicit accountability towards each other. The eﬀects of these control structures were unpredictable as organizational members responded diﬀerently to the introduction of these new rules that were enforced by the higher management of the member banks. The new control structures were the following: ● ● ●

Systems that attempted to administrate personal contributions. Evaluations on a personal level, attempting to confront people with their ‘production’. Financial evaluation schemes that act as an incentive. In Rabobank terms, this was called the ‘Employee Value Plan’.

These formal procedures were basically formal rules that were meant to induce new behaviours and, ultimately, new routines. But these rules had no foundation in existing behaviours, as employees had no scripts available to guide their interpretation of the new rules. The rules were not opposed or embraced, but they were for the larger part ignored. As such, they did not evoke the emergence of a control routine. The very limited impact of these new rules was also due to ambiguities caused by incompatibilities between various formal rules. People at diﬀerent banks were being evaluated on their contribution to the organizational goals according to a programme called Result Oriented Rewarding, a spin-oﬀ of Result Oriented Management. To make people personally accountable for their objectives was new to most member banks. However, as a respondent at a participating member bank remarks: The targets per employee, those are almost impossible . . . Now we are stuck with a system in which employees must tally their own production. And based on that, we have to determine if people are achieving their targets. So there isn’t a system in which a sales person enters the savings accounts he has just sold, and those sales get added to the status of that person. Rather, one needs to enter the customer in multiple systems, then you add something for your own file about

176

Organizational routines and stability in organizations

the customer, and then you need to establish your sales. So basically you can tally as much as you want. We do have guidelines, but some people will tally their activities sooner than others. So that is diﬃcult to measure. And now the bank is talking about performance related pay, but with the current systems, that is not really feasible.

Clearly, the formal rules that were introduced left an infinite amount of behavioural options open to the organizational participants. For every new rule introduced, more rules were needed to assist in the interpretation of the original ones.5 Related to the attempts to influence behaviour through performancerelated pay, a respondent from the West Zeeuws-Vlaanderen member bank noted: We do have the Employee Value Plan, but no-one is seriously considering this, except in March, when we learn how much we will receive. But no-one thinks: ‘If I try real hard the coming month, then there will be something extra for me next year in March.’ The Employee Value Plan, in my eyes it does not make any sense.

Indeed, at the announcement of the amounts, no one outside the management team made any reference to their performance in relation to amounts awarded. If we consider managerial action to be expressions of a desired formal state, then more sources of ambiguity were present: There has been much uncertainty, much change. It has been said that Client Advice needs to decrease in numbers of employees, now the project FAST is approaching so Financial Advice needs to decrease in size. It has been said that the fiscal harmonisation will be a threat to the bank and that Investor Relations needs to reduce its size, but nothing actually does!

The formal rules embodied in the accounting systems and those formal rules embodied in formal evaluations were not mutually consistent. The management of the various banks wanted to improve the accountability of its employees. These formally expressed rules of behaviour were to be enforced by specific accounting techniques, such as control systems and performance-related pay. However, both were unsuited to supporting the aspirations of the management team. The inconsistencies between these formal arrangements caused ambiguities, which led many participants to retain the routine behaviours they were familiar with. Moreover, through the absence of a history in which scripted interaction patterns could develop, there were very few, if any, practices that could support the formal rules of control in their early stages. As a result of the

Ambiguity between formal rules and routines of interaction

177

Formal arrangements

Rules: evaluation to enforce accountability

Routines

Tacit scripts Ambiguity Rules: accounting system indicates problem

Figure 8.4

Conscious scripts

Ambiguity between rules, eﬀectively excluding pressures for change in behavioural patterns

absence of these behavioural patterns and contradictory formal structures, people were not prepared to alter routine behaviours. Figure 8.4 illustrates that the formal requirement of control was problematic in its own right. The systems and formal procedures meant to support this requirement were not fully operational. As such, ambiguity was primarily caused by contradictions in formal signals. Therefore, the term ‘ambiguity’ is depicted in the box denoting ‘formal arrangements’. The link between scripts and formal rules is depicted here with dotted lines, to indicate that formal rules could not draw directly on existing scripted knowledge on control, since organizational participants had little experience with control in a day-to-day context. The result was very little influence on patterns of interaction, which could easily be construed as resistance (Prasad and Prasad, 2000).

DISCUSSION This chapter explored how inertia arose in the dialectic of new rules and routines. Through an emphasis on the interaction between rules and routines, a tentative characterization can be given of the role of inertia and ambiguity. Inertia, defined as ‘the inability to enact internal change in the face of significant external change’ is caused by a wide range of influences that inhibit organizational participants to enact organizational changes. We propose that one of these influences is ambiguity. Ambiguity can lead to various levels of resistance (Lowe and McIntosh, 2007) and we have identified a few

178

Organizational routines and stability in organizations

subtle illustrations in this chapter. In the Rabobank, two instances of ambiguity were identified which either resulted from rules that were incompatible with existing routines, or from rules that were mutually incompatible. The first instance of ambiguity arose from the failure of management and employees to interact as a result of diﬀerences in perceived meanings. The aﬀected employees maintained ontological security by interpreting formal signals through the routines in place. Thus, alarming figures from the accounting system had a profound eﬀect on the management team members, but they did not trigger similar reactions by the aﬀected employees. The second instance of ambiguity arose from the introduction of various inconsistent formal rules and an absence of related routines. The formal signals were so ambiguous that they had a limited eﬀect on the formation of new routines. Even more so, people turned more to other, existing routines, which actually gained validity as they proved to be unchallenged by the informal social setting of the employees (which was mostly outside the scope of managers). These occurrences of ambiguity led to practical problems at various member banks. Various banks reported an inability to mobilize broad commitment and support outside the project teams, which consisted of various organizational participants. The project leaders and consultants were often very pleased with the progress made in the project organization in terms of commitment and the achievement of formal goals. However, the achievements outside these project teams were often felt to be poor. As noted earlier, member banks strived for a bank-wide change in attitudes towards accountability. These were diﬃcult to achieve for the entire bank, because of ambiguities that acted as inertial forces. Every introduction of new rules, aimed at changing habitual behaviours, introduced some degree of ambiguity. In our analysis, ambiguities between rules and routines, as well as ambiguity between various rules resulted in inertia. The introduction of new rules did not constrain any of the organizational participants to any single behavioural alternative. Additionally, the execution of these rules was diﬃcult to monitor, as they generally referred to attitudes and preferred ways of interpreting and dealing with day-to-day situations. Therefore, in attempts to reduce ontological insecurity, many people were able to revert to existing behaviours, eﬀectively dismissing the new rules. A similar issue could be noted when various inconsistent rules were introduced. Many of these rules were not grounded in existing practices (a control routine was rather new), which made interpretations of these rules and their acceptance rather problematic. Again, organizational participants were able to retain existing routines, as the new rules were diﬃcult to monitor. Kripke quotes Wittgenstein by stating: ‘no course of action could be determined by a rule, because every course of action can be made to accord

Ambiguity between formal rules and routines of interaction

179

with the rule’ (1982, p. 7), which emphasizes the possibility that new rules can be interpreted in a manner that accommodates existing routines. As there is no unique interpretation of new rules, senior managers are usually empowered to make a decision on their proper interpretation. Managers use incentive systems and various monitoring systems in attempts to influence the dominant interpretation of new rules. A discussion of these systems would introduce an element of power into the discussion of rules and routines, which is an important topic in itself. But an additional complication is that systems designed to monitor and reward rule-following behaviour often consist of rules in themselves. As such, these systems introduce an additional level of complexity as they are also subject to interpretation.6 The use of rules to enforce other rules was problematic at various member banks as these new rules introduced additional ambiguities. We can therefore argue that rules do not necessarily impact routines directly as they may require additional rules to privilege some interpretations over others. Rather, rules must be seen in their context, which greatly aﬀects the degree of ambiguity they cause, and thus the degree of inertia that is associated with their introduction in day-to-day behaviours.

CONCLUSION The typology used by this chapter was aimed at the unravelling of one of the causes of inertia in routines and institutions: ambiguity. Ambiguity is a rather general term, but we argue that it can be located within the rules– routines dialectic. ROM, as a process of management accounting change, was very suitable for the study of this dialectic, because it entailed a well documented formal component in terms of management accounting rules, and it aimed to influence routine behaviours of organizational participants. By further dissecting routines into tacitly and consciously invoked scripts, we feel that it is possible to explore the relation between rules and routines on an individual level. Although this chapter has not provided an empirical discussion on tacitly invoked scripts and consciously invoked scripts, this separation allows for future attempts to do so. Intuitively, this makes sense, as tacitly invoked scripts will be less susceptible to change in response to new rules than consciously invoked scripts. This separation in two types of scripts may therefore reflect variation in the likeliness of ambiguity emerging in the rules–routines dialectic. Apart from this categorization of scripts, a second avenue for future enquiry is an extension of the source of inertia. This work has only discussed the role of ambiguity in the rules–routines dialectic, but it is highly unlikely

180

Organizational routines and stability in organizations

that inertia does not manifest itself in more ways, or is caused by additional processes. Insight into the various ways in which inertia impedes changes in routines and institutions may lead to a better understanding of the processes of organizational change and the role of formal rules in these processes.

NOTES 1. Institutions can be defined as: ‘the shared taken-for-granted assumptions which identify categories of human actors and their appropriate activities and relationships’ (Burns and Scapens, 2000: 8). 2. Becker refers to these types of patterns as the ‘contents of patterns’. 3. One can argue that the military relies on tacitly invoked scripts, for example when an aeroplane approaches an aircraft carrier. However, there is often a decision-moment present, where a sequence of scripts is consciously invoked. For example, when the aeroplane approaching is damaged, a decision can be made to invoke emergency scripts. These result in ‘automatic’ behaviours, but initially, the script is invoked consciously. 4. ROM management manual, 2nd version, pp. 6–7. 5. This is an example of the ‘Kripke–Wittgenstein paradox’, discussed by Reynaud (2005). 6. See the earlier example which argued that organizational participants had much discretion in tallying their contributions to total sales.

REFERENCES Balogun, J. and G. Johnson (2004), ‘Organizational restructuring and middle manager sensemaking’, Academy of Management Journal, 47(4), 523–49. Barley, S.R. and S.P. Tolbert (1997), ‘Institutionalization and structuration: studying the links between action and institution’, Organization Studies, 18(1), 93–117. Barnes, W., M. Gartland et al. (2004), ‘Old habits die hard: path dependency and behavioral lock-in’, Journal of Economic Issues, 38(2), 371–7. Becker, M.C. (2004), ‘Organizational routines: a review of the literature’, Industrial and Corporate Change, 13(4), 643–77. Becker, M.C. (2005), ‘A framework for applying organizational routines in empirical research: Linking antecedents, characteristics and performance outcomes of recurrent interaction patterns’, Industrial and Corporate Change, 14(5), 817–46. Becker, M.C. and T. Knudsen (2004), ‘The role of routines in reducing pervasive uncertainty’, Journal of Business Research, 58(6), 746–57. Berger, P. and T. Luckmann (1979), The Social Construction of Reality: A Treatise in the Sociology of Knowledge, Harmondsworth, UK: Peregrine Books. Burns, J.E. and R.W. Scapens (2000), ‘Conceptualising management accounting change: an institutional framework’, Management Accounting Research, 11, 3–25. Busco, C., A. Riccaboni et al. (2006), ‘Trust for accounting and accounting for trust’, Management Accounting Research, 17(1), 11–41. Cohen, M.D. (1996), ‘Individual learning and organizational routine: emerging connections’, in M.D. Cohen and L. Sproull (eds), Organizational Learning, Thousand Oaks, CA, USA: Sage Publications, pp. 188–94.

Ambiguity between formal rules and routines of interaction

181

Cohen, M.D. and P. Bacdayan (1994), ‘Organizational routines are stored as procedural memory: evidence from a laboratory study’, Organization Science, 5(4), 554–68. Cohen, M.D., R. Burkhart et al. (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5(3), 653–98. Dosi, G. (1997), ‘Opportunities, incentives and the collective patterns of technological change’, The Economic Journal, 107, 1530–47. Egidi, M. and A. Narduzzo (1997), ‘The emergence of path-dependent behaviors in cooperative contexts’, International Journal of Industrial Organization, 15(6), 677–709. Eisenhardt, K.M. (1989), ‘Building theories from case study research’, Academy of Management Review, 14(4), 532–50. Feldman, M.S. (2000), ‘Organizational routines as a source of continuous change’, Organization Science, 11(6), 611–29. Feldman, M.S. (2003), ‘A performative perspective on stability and change in organizational routines’, Industrial and Corporate Change, 12(4), 727–52. Feldman, M.S. and T.B. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48(1), 94–118. Feldman, M.S. and A. Rafaeli (2002), ‘Organizational routines as sources of connections and understandings’, Journal of Management Studies, 39(3), 309–31. Giddens, A. (1984), The Constitution of Society: Outline of the Theory of Structuration, Berkeley, CA, USA: University of California Press. Gilbert, C.G. (2005), ‘Unbundling the structure of inertia: resource versus routine rigidity’, Academy of Management Journal, 48(5), 741–63. Gioia, D.A. (1992), ‘Pinto fires and personal ethics: a script analysis of missed opportunities’, Journal of Business Ethics, 11(5–6), 379–89. Gioia, D.A. and P.P. Poole (1984), ‘Scripts in organizational behavior’, Academy of Management Review, 9(3), 449–59. Harris, S.G. (1994), ‘Organizational culture and individual sensemaking: a schemabased perspective’, Organization Science, 5(3), 309–21. Hodgson, M.G. and T. Knudsen (2004), ‘The firm as an interactor: firms as vehicles for habits and routines’, Journal of Evolutionary Economics, 14(3), 281–307. Kripke, S.A. (1982), Wittgenstein on Rules and Private Language: An Elementary Exposition, Cambridge, MA, USA: Harvard University Press. Labianca, G., B. Gray et al. (2000), ‘A grounded model of organizational schema change during empowerment’, Organization Studies, 11(2), 235–57. Lazaric, N. and A. Raybaut (2005), ‘Knowledge, hierarchy and the selection of routines: an interpretative model with group interactions’, Journal of Evolutionary Economics, 15(4), 393–421. Louis, M.R. (1980), ‘Surprise and sense making: what newcomers experience in entering unfamiliar organizational settings’, Administrative Science Quarterly, 25, 226–51. Lowe, A. and A. McIntosh (2007), ‘Knowledge management in a New Zealand tree farming company’, Journal of Organizational Change Management, 20(4), 539–58. Miller, D. and P.H. Friesen (1980), ‘Momentum and revolution in organizational adaptation’, Academy of Management Journal, 23(4), 591–614.

182

Organizational routines and stability in organizations

Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Harvard University Press. Nooteboom, B. (2000), Learning and Innovation in Organizations and Economies, New York, USA: Oxford University Press. Porac, J.F., G.R. Ferris et al. (1983), ‘Causal attributions, aﬀect, and expectations for a day’s work performance’, Academy of Management Journal, 26(2), 285–96. Prasad, P. and A. Prasad (2000), ‘Stretching the iron cage: the constitution and implications of routine workplace resistance’, Organization Science, 11(4), 387–403. Reynaud, B. (1996), ‘Collective dynamics: reflections on the introduction of a salary rule in a maintenance workshop’, Industrial and Corporate Change, 5(3), 699–721. Reynaud, B. (2005), ‘The void at the heart of rules: routines in the context of rule-following. The case of the Paris metro workshop’, Industrial and Corporate Change, 14(5), 847–71. Schank, R.C. and R.P. Abelson (1977), Scripts, Plans, Goals and Understanding: An Inquiry into Human Knowledge Structures, Hillsdale, NJ, USA: Lawrence Erlbaum Associates. Schein, E.H. (1985), Organizational Culture and Leadership, San Francisco, CA, USA: Jossey-Bass. Seo, M.-G. and W.E.D. Creed (2002), ‘Institutional contradictions, praxis, and institutional change: a dialectical perspective’, Academy of Management Review, 27(2), 222–47. Sharp, D.J. (1994), ‘The eﬀectiveness of routine-based decision processes: the case of international pricing’, Journal of Socio-Economics, 23(1/2), 131–48. Sluyterman, K., J. Dankers et al. (1998), Het Cooperatieve Alternatief, Honderd Jaar Rabobank 1898–1998, Den Haag, The Netherlands: Sdu Uitgevers. Teece, D. and G. Pisano (1994), ‘The dynamic capabilities of firms: an introduction’, Industrial and Corporate Change, 3, 537–56. Tripsas, M. and G. Gavetti (2000), ‘Capabilities, cognition, and inertia: evidence from digital imaging’, Strategic Management Journal, 21(10–11), 1147–61. van der Steen, M.P. (2006), Human Agency in Management Accounting Change: A Cognitive Approach to Institutional Theory, Ridderkerk, The Netherlands: Labyrinth Publications. Van Maanen, J. (1979), ‘The fact of fiction in organizational ethnography’, Administrative Science Quarterly, 24, 539–50. Weick, K.E. (1979), The Social Psychology of Organizing, Reading, MA, USA: Addison-Wesley. Weick, K.E. (1995), Sensemaking in Organizations, Thousand Oaks, CA, USA: Sage Publications. Weick, K.E. and K.H. Roberts (1993), ‘Collective mind in organizations: heedful interrelating on flight decks’, Administrative Science Quarterly, 38(3), 357. Woﬀord, J.C. (1994), ‘An examination of the cognitive processes used to handle employee job problems’, Academy of Management Journal, 37(1), 180–92. Yin, R.K. (2003), Case Study Research: Design and Methods, Thousand Oaks, CA, USA: Sage Publications.

PART IV

Organizational routines and organizational change and innovation

9.

The influence of artefacts and distributed agencies on routines’ dynamics: from representation to performation* Luciana D’Adderio

1.

INTRODUCTION

As a unit of analysis, routines represent an invaluable resource to capture organizational change (Simon, 1947; Cyert and March, 1963; Nelson and Winter, 1982; Becker et al., 2005; Pentland and Feldman, 2005a). Revealing the internal structure of routines can provide useful insights into many of the basic questions of Organization Science (Pentland and Feldman, 2005a). Yet, the complexity of this endeavour has meant that Routine Theory to date has only just begun to address the routines dynamics that underpin core organizational phenomena such as learning, change and adaptation. In particular, notwithstanding the important recent advances in this debate, we are still short of a full theoretical understanding and empirical characterization of the micro-level dynamics that underpin routines’ evolution. These include the dynamics of interaction between diﬀerent aspects of routines and the influence of artefacts and agencies on routines evolution. This gap in the theory has been exposed by authors who have advocated the need to unravel routines’ and capabilities’ internal dynamics (Pentland and Rueter, 1994; Cohen et al., 1996; Feldman, 2000; Lazaric and Denis, 2001; Zollo and Winter, 2002; Feldman and Pentland, 2003; D’Adderio, 2001, 2003; Becker et al., 2005; Pentland and Feldman, 2005a). This work has pointed to the need to ‘open up the routines black box’ to analyse the interactions between diﬀerent sides, or aspects, of routines. Categories introduced to capture the routines’ internal mechanisms include the distinctions between ‘routines-as-representations’ and ‘routines-as-expressions’ (Cohen et al., 1996), ‘rules-to-be-interpreted’ and ‘rules-to-be-executed’ (Reynaud 1996, in ibid.) and ‘ostensive’ and ‘performative’ (Feldman and Pentland, 2003; Pentland and Feldman, 2005a). These approaches have

185

186

Organizational routines and organizational change and innovation

productively shifted the emphasis from a characterization of routines as undiﬀerentiated monolithic ‘objects’ to the more sophisticated and productive notion of routines as generative – and continuously emerging – systems characterized by internal structures and dynamics. In so doing, authors have opened up entirely new grounds for exploring some of the most relevant but as yet under-researched questions about the nature and dynamics of routines. This chapter aims to contribute to fill this gap in our understanding of routines’ evolution and performance. In our quest to unravel routines’ internal dynamics we set our focus on artefactual representations of routines – and, specifically, on standard operating procedures (SOPs) and associated rules – which we use as a starting point for our analysis.1 There are two ways in which a focus on artefacts is useful. First, artefactembedded rules and procedures provide vantage points to observe the ostensive (abstract) aspects of routines with respect to which they can serve as ‘proxies’ (Pentland and Feldman, 2005b). When embedded in material artefacts, rules and procedures can provide ideal loci for observing abstract understandings and otherwise embodied views of routines; this is because they become more stable and visible, which in turn allows them to act as reference points against which variations occurring in performances can be more easily detected.2 The second way to understand the key role of artefacts as privileged points of observation – which is perhaps more akin to the framework developed here – is that abstract understandings of routines are not simply people-embodied but highly distributed across a complex web of people and everyday artefacts.3 Neglecting to include tools and artefacts in the study of routines’ dynamics can only provide at best a partial picture. Starting from these premises, we identify three main research questions. First, how can we theorize about the mutual adaptation of formal (artefact-embedded) routines and rules and actual performances? For example, what are the micro dynamics that influence the direction and intensity of their interactions, and how do these dynamics influence routines’ evolution and adaptation? Second, what is the role of artefacts in mediating these interactions? Third, what is the influence of distributed agencies, including heterogeneous organizational communities, in shaping the co-evolution between diﬀerent aspects of routines? To answer these questions we need to acquire a new set of theoretical notions and constructs. Drawing from recent arguments within Economic Sociology and the Sociology of Financial Markets (Callon, 1998, 2007; MacKenzie, 2003, 2006a, 2006b), we theorize the interactions between formal SOPs and rules, on one hand, and performances, on the other, as iterative cycles of framing, overflowing and further reframing of knowledge inputs and actions. The framing action exerted by SOPs, rules and

The influence of artefacts and distributed agencies on routines’ dynamics 187

formal tools delimits and closes search spaces, providing guidance and control. Framing by rules and SOPs, however, is never complete: there is always overflowing which opens up search spaces, thus introducing scope for divergence, adaptation and change. Overflowing is often followed by further reframing, which again brings convergence between the procedure(s) or rule(s) and performance(s). The combined novel empirical focus and theoretical characterization provides important new insights into the mutual adaptation between aspects of routines and the role of artefacts and other agencies as intermediaries in these interactions. Specifically, our framework provides three main contributions to the routines debate. First, we are able to capture the interactions between diﬀerent sides of routines: the cycles of framing, overflowing and reframing – and their emergent outcomes – form the micro dynamics underpinning the mutual adaptation of diﬀerent aspects of routines and, ultimately, routines’ evolution. Second, we characterize the influence of artefacts as intermediaries in shaping the interactions between diﬀerent sides of routines. In contrast with extant literature that sees formal SOPs and rules as either flawed representations that can be easily dismissed/disused or as full prescriptions that are compulsively and automatically performed, we show that – in most cases – there is some kind of adaptation. Third, we theorize the influence of distributed, heterogeneous agencies (that is occupational communities, communities of practice) on routines’ evolution. We show how routines’ dynamics are the emergent result of contingent struggles amongst competing performative programmes. Our data was obtained through the ethnographic observation of the ‘engineering freeze’ process, an upstream section of the product development process at a leading automotive manufacturer. Here, the introduction of a software-based data and process management tool (Product Data Manager, or PDM) and the consequent inscription of the ‘freeze’ routine in software provide vantage points to observe the interaction and mutual shaping of SOPs and performances. The data collection, aimed at documenting the actual contents of routines (Cohen et al., 1996), involved a mix of participant-observation and in-depth, semi-structured interviews. These were conducted over a one and a half year period and spanned most organizational functions (Industrial Design, Product Engineering, Analysis, Testing, Production, Manufacturing, Accountancy, Marketing) and levels (designers, shop floor technicians, project and programme managers). The author’s technical background allowed for full immersion into the micro-level interactions among processes, artefacts, people and technologies. The data was subsequently analysed and compiled into a case study consistent with an inductive, ‘grounded theorizing’ approach (Glaser and Strauss, 1967; Eisenhardt, 1989).

188

Organizational routines and organizational change and innovation

The chapter is structured as follows. Section 2 sets out the main theoretical and conceptual foundations for the three main themes: the interactions between aspects of routines and the influence of artefacts and agencies on these interactions. Section 3 outlines the empirical context, by describing the changes occurring to a key product development routine (the ‘freeze’ process) as a consequence of software introduction. The main body of evidence is presented in section 4. Section 5 discusses the insights that can be obtained by applying Performativity Theory to the study of routines. Section 6 concludes by outlining the implications for Routine Theory.

2.

ROUTINES, ARTEFACTS AND AGENCY: FROM REPRESENTATION TO PERFORMATION

In this section we lay out the theoretical premises to the framework that we propose to use to capture the mutual adaptation between artefactual representations of routines (rules, SOPs) and performances, and the influence of agencies on their interactions.4 Our aim is to argue for a shift away from two extremes that dominate the routines debate: a characterization of SOPs and formal rules as fixed representations of actual process that fully prescribe actions; and a characterization of SOPs as intrinsically flawed, merely descriptive representations of actual process that can be easily avoided. In turn, we put forward a ‘performative view’ of routines dynamics.5 This captures the interactions between SOPs and performances as iterative cycles of framing, overflowing and reframing, and highlights the fundamental role that artefacts and distributed agencies play in shaping their interactions. 2.1

Routines, Artefacts and the Role of Agency

In our quest to characterize routines’ internal dynamics, we thus begin by focusing on the interactions between formal – often artefact-embedded – rules and procedures, on one hand, and actual performances, on the other. When we consider the key role that SOPs and rules play in the emergence and evolution of routines, the scarce attention that they have received in the literature, since Cohen et al.’s 1996 seminal paper, is rather surprising.6 For routines scholars, this represents a missed opportunity to capture a fundamental issue in routine dynamics: while artefact-embedded representations of rules and routines are mostly introduced to design and manage routines, their outcomes often escape the agents’ original intentions. In other words, artefactual representations of routines are not the routine (Pentland and Feldman, 2005b).7 This, however, should not deter us from

The influence of artefacts and distributed agencies on routines’ dynamics 189

using artefactual manifestations of routines as starting points for our analysis. Rather, the processes of translation from SOP to performances and vice versa (coding and de-coding) provide very interesting standpoints to observe the production and reproduction of routines (Cohen et al., 1996; D’Adderio, 2003).8 For example, to what extent do explicit, formal rules and procedures govern practice? How are procedures and their outcomes shaped by contingencies? What are the mechanisms that regulate the mutual influence and adaptation between artefact-embedded rules/ procedures and performances? 2.1.1

The influence of artefacts and agencies on routines evolution: description vs. prescription When considering the influence of formal rules and SOPs over performances, we can identify two contrasting schools of thought: a ‘framing’ view, which focuses on the power of objective structures to define, prescribe and frame actions; and an ‘overflowing’ view, which focuses on the ability by human agents to interpret, modify and, in some cases, fully override a rule or procedure. The ‘framing’ view, often also referred to as the ‘rationalist’, ‘cognitivist’ or ‘mechanistic’ view (cf. Tsoukas, 1996) is common to much technical (AI, computer science) and positivistic management literature (since Taylor’s 1911 classic contribution). This view sees rules and procedures as fixed and objective representations which reflect accurately the human practices that they are designed to prescribe. According to this view, thus, formal procedures and rules, as reified, external representations, univocally determine the course of human practices; they are reproduced automatically and diﬀuse linearly. This understanding at the extreme sees rulesas-representations as causally operative, thus providing a distortion that mistakenly conflates the rule or formula with its enactment (cf. Bourdieu, 1990, see also Taylor, 1993). Human actors are reduced to just ‘cogs in the wheel of a larger technical system’, rule-following automata that reproduce a fixed routine for a fixed outcome (Berg, 1998: 467). In this case, thus, there is no distance between the rule or procedure as represented and the rule in eﬀect: the rule or procedure determines the practice. The ‘overflowing’ view provides a valid critique to the mechanistic view of rules and procedures portrayed above. Where the framing tradition emphasizes the purity of logic-driven, automatically reproducing rules, the ‘overflowing’ view emphasizes by contrast the intrinsic complexity, variety and adaptability of human practices as well as the power of discretion by human actors to interpret, modify and even completely reject a rule or procedure. Widespread in Sociology, Ethnomethodology, and, increasingly, in Organizational Studies, this view often portrays the ‘objectivist’

190

Organizational routines and organizational change and innovation

representations of procedures as external entities as selective, at best incomplete and at worst fundamentally flawed reproductions of ‘real’ processes which they attempt to capture, imitate, guide or direct often with only partial success (Lynch, 2001). As ‘no representation of the skills involved in performing appropriate human activity can ever be adequate [. . .] the ability of formulations to guide what people do rests on abilities to use and understand them’ (ibid.: 8–9, emphasis added). Views animated by Wittgenstein’s philosophy (including early social constructivists and ethnomethodologists), for example, have highlighted how structures, including rules and classifications, are never deterministic but always interpreted (Garfinkel, 1967; Barnes, 1982; Bloor, 1973, Hatherly et al., 2007; Lynch, 1992). The irreducible interpretive flexibility of rules is such that – at least in theory – it can lead to infinite regression as ‘no course of action could be determined by a rule, because every course of action can be made out to accord with the rule’ (Wittgenstein, 1967: 81). Due to the logical under-determination of behaviour by rules, there can be no closure as ‘The rule is, at any given time, what the practice has made it’ (Taylor, 1993: 57–8 emphasis in original).9 In other words, the practice determines the rule. Organizational scholars have drawn on these approaches in conceptualizing the role that technology plays as a source of constraint in rule-following (Barley, 1986; Orlikowski, 1992). Also in this case, however, the emphasis remains firmly on the discretion by human agents to follow a procedure or rule or ‘choose to do otherwise’ (Giddens, 1993; Orlikowski, 1992). According to this framework, the properties embedded inside artefacts are never predetermined but rather ‘the capacity to modify the “rule” that is drawn on in any action is an ever present possibility’ (Cassell, 1993: 13, see also Orlikowski, 2000). Rules in this view exist only virtually and are consequential only to the extent that they are enacted by users through practice. The inherent flexibility and adaptability of human practices implies that rules may attempt to guide behaviour (Spender, 1989) but human actors can always operate discretion in interpreting the rule or procedure, assign meanings (Daft and Weick, 1984) and ultimately decide whether, how and when to abide, work around, or altogether reject them.10 To summarize, while the contribution of the ‘overflowing’ view is fundamental in dispelling technological determinism in showing how rules are interpreted and enacted in the context of actual practice, it tends to overemphasize the power of human agents’ discretion. At the extreme, rules and procedures become consequential – they have an eﬀect over reality – only when interpreted and/or enacted by humans. Thus, while according to the framing view, formal SOPs and rules are mainly prescriptive, according to the overflowing view, they can become merely descriptive as actors can

The influence of artefacts and distributed agencies on routines’ dynamics 191

always choose to avoid them. By contrast it is our contention that, while there is always scope for human intervention, formal rules and procedures have a more fundamental influence on rule-following than is allowed by the latter view. This is especially true when rules and procedures are embedded in technological artefacts. 2.1.2 Rule-following as distributed cognition and the role of artefacts While in rule-following as well as in other realms of human practice, artefacts do not determine actions, nevertheless they ‘plainly matter’ (Hatherly et al., 2007). Technical systems, for example, make it possible or easy to do certain things, and impossible to do others, so that, while in theory there can be infinite regression, ‘the logical open-endedness of the application of terms to particulars and the logical under-determination of behaviour by rules are foreclosed in practice’ (ibid.: 11, emphasis added). Science and technology studies authors, for example, have shown how a complex range of rules and assumptions as ‘scripts’ are embedded within technology both at the design and usage stage (Akrich, 1992; Latour, 1992, Grint and Woolgar, 1992).11 This involves a socio-technical process of ‘inscription’ (Latour, 1992) by which dominant interests are reflected in the form and functioning of a technology. The existence of technology-embedded rules implies that focusing solely on human actors to explain rule-following practice is inadequate. Rule-following, as depicted in Ethnomethodology and Cognitive Anthropology, is in fact a highly distributed set of knowledge and activities which stretches across a mutually supportive array of elements including material devices, language modes and representation modes (Lynch, 1995 in Preda, 2000). In his ethnography of the two diﬀerent scientific practices of ‘opticism’ and ‘digitality’, for example, Lynch shows how rule-following depends on the relationships among material devices, theoretical optics, geometry and modes of graphic representation (ibid.). Similarly, Hutchins (1995) showed how the process of piloting a ship in and out of a harbour is a complex, rule-determined activity involving not only the coordination of crew members, but also the use of navigational instruments and maps. A gyrocompass, for example, incorporates some essential rules of sea and land orientation and for this reason it requires that the pilots’ skills and activities be adjusted to its properties (Hutchins, 1991). Once embedded in artefacts, skills and tacit knowledge (Latour, 1992; D’Adderio, 2001), rules (Hutchins, 1991; Preda, 2000; D’Adderio, 2003), and procedures (D’Adderio, 2003; Hatherly et al., 2007) tend to become more stable and durable which holds radical implications for rule-following behaviour.12 The role of software We focus on a category of artefacts which is both very interesting and relevant: software. Information systems, as bundles

192

Organizational routines and organizational change and innovation

of inscriptions, play a fundamental role in influencing rule-following. Such systems are ‘neither merely neutral media nor simply means of increasing the eﬃciency of what unaided human beings might do’ (Hatherly et al., 2007: 32).13 They structure work, extend interactions, increase visibility of knowledge and actions, create a common platform for the accumulation of common knowledge, constrain the ability of practitioners to alter the results of another, regulate who has access to making changes, track progress of changes, link multiple sites in diﬀerent time and geographical locations, facilitate data sharing and the reception of feedback (Orlikowski, 2002). They solidify and stabilize rules, procedures and classifications, thus making it more diﬃcult to avert them (Hatherly et al., 2007). While technical constraint is never absolute and indeed many system controls can be subverted if suﬃcient resources and incentives are applied to the task, there are several reasons why the influence of technologies – in general – and information systems – in particular – is critical. While in theory it is always possible to bypass software-embedded controls, in practice this does not always occur. A first reason is that assumptions, rules, procedures and classifications, embedded in software both at design and usage stages, tend to sink in and become part of the users’ habitual background (Bourdieu, 1977), or ‘the way we do things around here’ and as such are often unquestioned. Second, as distributed and pervasive technologies, information systems are often entangled into a thick web of organizational relationships which makes them diﬃcult to avoid. Once adopted, for example, software can influence what kind of information should be created, selected and shared, with whom, in which format and in what sequence. While practitioners can often choose to bypass the software, their boycott will hold consequences for them in terms of their ability to have their actions or feedback taken into account by others in the organization. Third, software tends to make information more visible across an organization, thus making it easier to control that actions actually comply with the software. A fourth and final reason is that, while formal software controls can in theory be easily modified or averted, in practice this requires the deployment of resources (that is time and programming competences) which are often not available.14 In these circumstances, the ‘power of default’ (Koch, 1998 in D’Adderio, 2003; Pollock and Cornford, 2004) will prevent adaptation and customization. The role of technological systems in influencing rule-following thus deserves to become a crucial topic. 2.2

From Representation to Performation

In our quest to characterize the influence of SOPs and rules over performances, we are therefore proposing a shift of emphasis from an objectified

The influence of artefacts and distributed agencies on routines’ dynamics 193

and detached view of rules and procedures as external objects that have fixed properties, to a performative view where rule-following is characterized as a typically emergent, distributed and artefact-mediated activity. Such a shift does not however correspond to abandoning previous approaches: performation includes representation, but recognizes that representation is only half the story. As ‘proxies of the ostensive’ or ‘representations’, SOPs and rules reflect – to an extent – abstract views, theories and imperatives. At the same time, however, SOPs and rules as statements are not alienated from but an intrinsic part of actual processes over which they exert substantial influence. Introducing performation allows us thus to explore the reciprocal influence of SOPs, rules and performances and can thus be read as ‘an act of rebalancing’ (Pickering, 1994). 2.2.1

Relationship between theories and the reality they model: disentanglement, performativity and counter-performativity The notion of performativity explains how theories and models are not merely simple descriptions of settings but transform the settings that they describe. Economic theory, thus, according to Callon, does not simply describe but it performs, it alters actual markets (1998, 1999). Similarly, in the context of industrial rule-following, SOPs and rules as models and theories of actual processes, alter the course of actual practices wherever they are introduced. In providing a framework to study the mutual adaptation of models of processes and actual processes, performativity theory can thus provide us with a novel and promising way to improve our characterization of rules and routines dynamics. To explain what is meant by performativity and how this can help us in our quest to characterize the dynamics of routine- and rule-following we draw from the work of Callon and MacKenzie on the performativity of financial markets theory. Building on the anthropologists’ notion of ‘entangled objects’ (Thomas, 1991) economic sociologist Michel Callon portrays the construction of economic markets as involving processes of ‘disentangling’, ‘framing’ and ‘overflowing’ (1998, 1999). Callon starts from the premise that the market, as a method of coordination, implies the existence of agents capable of calculation. To make ‘calculativeness’ possible, however, specific conditions must be put in place. If calculations are to be performed and completed, the agents and goods involved must be ‘disentangled’ and ‘framed’. Framing involves the drawing of ‘a clear and precise boundary between the relations which the agents will take into account and which will serve in their calculations and those which will be thrown out of the calculation’ (1998: 16). Particular attention here is devoted to the role of tools, equipment and devices which contribute to the framing of transactions. Tools are mediators between the theory of economics and the economy: ‘not only are they

194

Organizational routines and organizational change and innovation

responsible for the cross-relations between the two but, like any other mediator, they promote the construction and constitution of each of them’ (Hennion, 1993, see also Callon, 1998: 28). Recalling Garcia’s (1986 in Callon 1999) example of the construction of the table strawberry market, Callon shows how tools and devices such as the display of transactions on the electronic board and the qualification of batches of strawberries on data slips were fundamental in giving the agent’s action shape, thus creating an arena or ‘space for calculability’ (1999: 191). This first movement, or ‘framing’ identifies a process of convergence between the model and the economy. Framing, however, is never complete: any frame is necessarily subject to ‘overflowing’, indicating a divergence between the model and the actual market. This is often followed by further ‘reframing’ by which there is again convergence between the model and the actual markets. Economic theories and models are performed through these iterative cycles of framing, overflowing and reframing which regulate their mutual adaptation. An important feature of a performative view is that economic models and theories are not ‘external’ to the market but an intrinsic part of it. There is in fact no real separation between ‘market models’ on one side, and ‘market practice’ on the other: market models are performed in practice. Models form a crucial part of markets; they are not purely detached external representations or virtual abstractions (cf. Miller, 1998 in Holm, 2002) but engines that make the markets tick. Similarly, while process models can be seen to an extent as blueprints of actual practices, or ‘the bit of sour dough that is used as a starter for the next loaf of bread’ (Pentland and Feldman, 2005b: 5), they are not separate from those practices. MacKenzie builds on Callon’s notion of performativity with his study of the market for financial derivatives (2003, 2006a, 2006b). He too shows how models are not simply a description of something resting outside the market (reality) but a constituent part of it. This is an important advance in our understanding of the role of models: these are not just passive ‘guiding principles’, setting the boundaries of what can be done and what can’t be done, as scholars have argued so far, but they contribute to shaping actual processes. MacKenzie’s work is especially of interest in his finer grained identification of diﬀerent categories of performativity or of influence of models on reality: ‘generic performativity’, when an aspect of economics (a theory, model, concept, procedure, dataset, and so on) is simply used by participants in economic processes; ‘eﬀective performativity’, when the practical use of an aspect of economics has an eﬀect on economic processes; ‘Barnesian performativity’, when the practical use of an aspect of economics makes economic processes more like their depiction by economics; and ‘counter-performativity’

The influence of artefacts and distributed agencies on routines’ dynamics 195

when the practical use of an aspect of economics makes economic processes less like their depiction by economics. This classification is especially useful as it leads to a much finer grained and therefore insightful characterization of the interactions between procedures/rules and performances. In particular, MacKenzie’s framework shows that what previous theories considered the norm, are in fact often exceptions. At one extreme of performativity, thus, there is prescription. Prescription represents a very strong instance of performativity: automatic reproduction, pure repetition, no more recalcitrance, recurrent events (Sahlins, 1985, see also Callon, 2007). Full prescription corresponds to the case of ‘fiat lux et lux fuit’, as illustrated by an automatically reproduced sequence of computer algorithms. At this extreme, which corresponds to the ‘framing view’ outlined above, there is very little or no adaptation, as models are automatically reproduced. At the other extreme, there is the full demise, rejection or disuse of a model or tool. This case corresponds to the ‘overflowing’ view outlined above: the influence of the model is so weak that it is bypassed, worked around or outright rejected and therefore is not enacted in practice. One way to explain the demise of a tool of course is that individual agents have made the conscious choice to reject the model. Performativity theory, however, while not denying this possibility, aﬀords us a more interesting explanation: the model as statement has not been able to put into motion a world in which it can function. In other words, the statement or formula has not been able to produce a successful socio-technical agencement.15 While full prescription and mere description are always possibilities, most of the time (and especially in the case of complex organizations operating in conditions of uncertainty) there is performativity, implying some kind of dynamic adaptation between model and reality (Callon, 2007). In conclusion, we postulate that the performativity framework can be eﬀectively harnessed to improve both our theoretical understanding and empirical characterization of the interactions between procedures and performances. It will also provide new grounds to characterize the key role of artefacts and tools in general – and software in particular – in mediating these interactions.

3.

INTERACTIONS BETWEEN PROCEDURES AND PERFORMANCES AND THE ROLE OF SOFTWARE

We thus focus our attention on software-embedded SOPs and observe their interactions with performances as they unravel in practice. For our analysis we have selected the ‘BoM (Bill of Materials) freeze process’, a

196

Organizational routines and organizational change and innovation

crucial segment of product development corresponding to the handover of a product configuration from Engineering to Production. In this context we examine the mutual adaptation between the computer-embedded freeze procedure (and related rules) and the actual process. Being so stable – and therefore easy to observe – the software procedure provides us with a useful device to capture the co-evolution of SOPs and performances. The fieldwork involved a one-and-a-half year ethnographic study based on participant observation at a leading automotive manufacturer. The data collection focused on a complex vehicle development programme – including one hundred vehicle variants – which was monitored for its entire duration (18 months). During this time the researcher was able to access the firm’s facilities with daily frequency thanks to the provision of a contractor’s ID badge. While being given a desk in the Advanced Technology and Product Development (ATPD) department, responsible for the implementation of leading-edge enterprise software technologies, the badge allowed the researcher to circulate freely throughout the organization. As a member of the ATPD implementation team she was also invited to take part in company-wide meetings, seminars, workshops and away days which provided valuable opportunities to discuss early findings and make informal acquaintance with potential interviewees from most organizational functions and levels. Evidence-gathering activities involved the direct observation of product development practices as they unravelled. It also involved recording practitioners’ accounts of practices through in-depth semi-structured interviews with programme managers, directors, product and engineering administrators, design and manufacturing engineers, industrial designers, marketing, sales and accounting personnel. Additional evidence was gathered through searching the company library and electronic archives, as well as scanning the many manuals, documents and databases that were created by the team and shared throughout the development process (up to manufacturing release). Less frequent update visits were carried out for a further period of 18 months following the start of production. The combination of direct observation and interviews over a significant time span provided excellent grounds to capture the interplay of procedures and performances, as well as the role of software as an intermediary in these interactions.

4.

THE ‘FREEZE’ PROCESS

The evidence on which our analysis is based draws from the in-depth observation of the Bill of Materials (BoM) freeze process (from now on the ‘freeze process’), a critical sub-segment of product development. Our

The influence of artefacts and distributed agencies on routines’ dynamics 197

example focuses on the changes occurring to the freeze process following the introduction of Product Data Manager (PDM) software upstream, at the Engineering end of product development.16 Here, the introduction of software provides a unique opportunity to capture the relationship between the formal freeze SOP (embedded in software) and actual performances as enacted by practitioners in our automotive firm at the time of fieldwork. 4.1

The Formal Process

We begin with the detailed account of the freeze process as described by a senior engineering administrator and recorded by the researcher during observation.17 Our engineer takes us through the complexities of the freeze process and the changes that are occurring as a consequence of software implementation. After highlighting the diﬀerence between the ‘formal’ and the ‘actual’ freeze process, our engineer begins to describe the formal freeze process, which is embedded in software as standardized ‘best practice’ (Figure 9.1). The formal process starts with the definition of the vehicle specification. This is driven by the ‘Test Plan’ whereby testing requirements are optimized against the vehicle to be produced and its variant configurations. The ‘Prototype Build Specification’ (PBS) document is then drawn, which states the number of prototypes that will be required for a specific vehicle, or vehicle family. Next, the prototype vehicles’ characteristics are specified Testing Test plan Optimize testing

Engineering

ERS

Vehicle spec requirements PBS (Prototype Build Spec)

Specify variants build tick-list Batch-1

‘Connect up’ variants to vehicles

PDS (Product Description Summary)

PPS (Product Programme Submission)

Source:

Interview/MC.

Figure 9.1

The prototype Bill of Materials (BoM) process

BoM freeze (no ECOs after this date)

ERS creates n × M/Views (one for each vehicle)

198

Organizational routines and organizational change and innovation

according to the ‘Product Description Summary’ (PDS) or marketing spec, which is a description of the standard vehicle to be produced. This can be driven by customer requirements, industry trends, benchmarking assessments, or by technology (that is testing) requirements. The PBS, or vehicle spec, is then passed on to Engineering Release Systems (ERS), where process administrators build a matrix-based tick-list, which identifies and lists all available vehicles’ options and their variants. After this, the vehicle specification returns to Engineering where it is divided up into batches. Taking into consideration one vehicle batch at a time, Engineering fills in the tick-list matrix by identifying which variants are to be included in each vehicle specification. Subsequently, ERS ‘connects up’ the variants to the vehicle. ‘Connecting up’ involves the creation of a structural relationship between the variants and the vehicle. The formal process at this point states that, once the volume of changes has diminished substantially, ERS should call the Bill of Materials freeze (the date after which only small changes are permitted) and proceed to create a number of individual configurations in the software’s ‘Manufacturing View’, one view for each vehicle variant configuration. This is where the formal and actual processes begin to diverge as detailed in the practitioners’ account of the actual process. 4.2

The Actual Process

The proportion of uncertainty and change that an organization faces varies substantially across diﬀerent stages of Product Development (Figure 9.2). The early stages are characterized by high levels of change cycles and design iterations with product parts and assemblies undergoing very frequent and radical modifications. Changes at this stage are implemented by raising an ‘Engineering Change Order’ and can be either initiated by a designer who is seeking to improve a part’s design or performance, or they may originate as a consequence of other changes aﬀecting a related part. The iteration cycles continue until the product structure and parts are completely defined and require only minor (or no further) alterations. The SOP at this point dictates that Programme Management and ERS should call the ‘freeze’. The freeze implies that the product data and structure are suﬃciently stable for the configuration to be released to production and manufacturing. Ideally, no more Change Orders should take place after the freeze milestone, because these would cause disruption to downstream development functions. While some flexibility is allowed during the early stages of development to enable Industrial Designers and Product Engineers to experiment by trial and error with alternative solutions for product parts and assemblies, after the freeze, control is mandatory to

The influence of artefacts and distributed agencies on routines’ dynamics 199

changes A B

time Source:

Interview/NG.

Figure 9.2

Incidence of changes along the product development life cycle

BoM Freeze

Buy/make parts

Define vehicle spec Start

Source:

Phase 1 Proto

Engineering Change Orders (ECOs)

Manufacturing Deviations

ENGINEERING E/View (as-designed)

MANUFACTURING M/View (as-built)

Interview/MC.

Figure 9.3

‘Request to change’ documents

allow for the stabilization and validation of product definition, which is required for optimizing Production tools and Manufacturing processes. To support the control of changes after the freeze, the software instigates the transfer of the product configuration from one single Engineering View (E/View) to multiple Manufacturing Views (M/Views) (Figure 9.3). The M/ Views are generated by loading up individual product variant configurations

200

Organizational routines and organizational change and innovation

on PDM. The diﬀerence between the E/View and M/View is that, in the former, changes must be implemented only once to be automatically propagated wherever the part is used; this is because in the E/View there is only one comprehensive structure from which all vehicle variants can be derived. All vehicles stored in the M/View, instead, are diﬀerent as each one corresponds to a specific vehicle variant; every change introduced after freeze must therefore be manually duplicated across each of the vehicle variant configurations, one at a time. This means that, from the freeze onwards, the vehicle configurations have to be maintained and modified individually, according to Manufacturing Change Requests (or Deviations) as and when these come through. By taking control of the transfer of the product configuration from the E/V to the M/View, therefore, software makes changes diﬃcult and cumbersome to implement, eﬀectively slowing down and making the process of change implementation ineﬃcient, thus favouring stability. This is in fact the rationale for calling the ‘freeze’: to reduce the number of changes and the extent of their fluctuations, therefore providing stability to downstream development functions. According to our engineering administrator, however, this is the point where the actual process begins to diverge from the ideal process. While – in theory – after the freeze the Release function should create one manufacturing view for each vehicle variant, the actual process is less straightforward, as the detailed observation of the X100 Vehicle Programme (ongoing at the time of fieldwork) illustrates. The X100 is a complex vehicle programme due to the high number of vehicle variants planned and the related high level of data, assemblies and configurations that have to be created and maintained. By the time X100 has reached the freeze milestone deadline, many parts have not yet been released, and a huge volume of change is still required before any prototypes can be manufactured: ‘At this stage we are still being deluged with a substantial number of Change Orders’ (Interview/MC). The process is still far from the nominal conditions required to call a freeze; the product structure is still highly unstable, the number of changes is high, and most of the changes required are still major. A project milestone date has been reached, where ERS must ‘freeze’ the Bill of Materials and move the product variants’ configurations from PDM’s E/View to the M/View in order to facilitate the control of manufacturing deviations. The development team, however, is still lagging behind, struggling to manage the enormous amount of data and assemblies that have been generated. The production-oriented logic embedded in PDM at design stage is intentionally devised to control and inhibit the introduction of changes after the freeze in order to stabilize the product configuration early to the benefit of Production and Manufacturing.18 PDM, however, works on the assumption that no major changes are required after the freeze, a situation

The influence of artefacts and distributed agencies on routines’ dynamics 201

that does not often occur in practice, especially in complex development programmes such as the one analysed here. Given the substantial amount of Change Orders that are still required to the first batch of vehicles at the time of freeze, and following the realization that the migration to the M/ Views would make it much harder to implement later changes, upstream engineers decided not to create the M/Views as required by the formal process and rules embedded in software: Theoretically, engineers are not allowed to attribute a ‘Batch-1’ eﬀectivity status to an Engineering Change Order, they have to do it on a Deviation. However, most of the deviation requests we are receiving today are saying: ‘Please can you apply this Order into Batch-1’. Today we are beyond freeze date and yet most requests concern Orders with Batch-1 eﬀectivity. It is clear that, if we can process that change into the E/View, as an ordinary Change Order, we need only to do it once. So that’s the main reason why we kept in the Engineering View. So that’s what we are doing today, we are processing the Order into E/View with Batch-1 eﬀectivity. (Interview/MC)

Overwhelmed by the amount of change, the team has decided to remain in the E/View. ‘Today, although we are after the freeze, we are keeping everything in the Engineering View. In the E/View you only need to make a change once, and it is automatically reflected everywhere the part aﬀected by the change is used’ (ibid.). Given that the incoming deviations for specific vehicles are so few at the time of the freeze, while the number of Change Orders is still very high, the decision to bypass the software embedded rule and keep everything in the Engineering View facilitates the implementation of changes. The workaround is completed by introducing a sub-procedure which consists in attaching a Deviation document to each modified part. Because the change is implemented in the E/View, every time they load data from this into any of the Manufacturing Views, or every time they look at the data in the E/View, the Deviation document will always appear in association to the part. This way, the material specification Deviation associated with the part can always be visualized, even though the oﬃcial manufacturing views have not yet been created. The decision to work around the software rule, however, is not without consequences. The formal rule dictating that variant configurations must be loaded in PDM’s M/View is aimed at enforcing greater control over individual changes and on the way these aﬀect each individual configuration. M/Views, for example, require an engineer to specify exactly from which view to which the change is to be propagated. For example, PDM will create an error every time one tries to propagate a change to a configuration where the part is non-existent. The software makes it necessary to be very specific as to where each part aﬀected by the change is used, and to which views a specific change is to be propagated: ‘This is due to

202

Organizational routines and organizational change and innovation

the structure of PDM’s logic, which is of an incremental nature, meaning that the software will allow only incremental changes and make any other type of changes very diﬃcult to implement’ (Interview/MC). The underlying philosophy of Product Manager and many other engineering control systems is in fact one of sequential incremental change: In PDM you have to talk in terms of ‘increment’, you have to specify exactly what it is that you are going to change. . . . In other words, you have a starting point today, and you can only change today’s data. So you make an incremental change, which gives you a new starting point, then you can make a further incremental change, then a further incremental change, etc. but they are all based on what has happened before. That gives you very clear control over the changes that you then send to production. (Interview/MC)

PDM operates a selective control action by allowing or disallowing specific actions and by facilitating or preventing specific changes to be implemented at diﬀerent times during the development process. While the production-oriented philosophy embedded in PDM is aimed at providing better control and supporting the implementation of changes, it is at times perceived as an unnecessary constraint and source of rigidity. In our case, to avoid such rigidity, the software procedure is bypassed: the decision is taken to postpone the creation of the M/Views until the time when the number of Deviations that are required for a specific vehicle will eﬀectively and substantially outweigh the volume of Engineering Orders. Workarounds in computing and other technologies are well documented in the technical, CSCW19 and STS literature (Gasser, 1996; see Pollock, 2005 for a review). However, authors in these traditions have not suﬃciently emphasized that, while it is often possible to bypass the software-embedded rules, this holds important systemic implications at the level of the organization. While working around a software procedure or rule may be feasible, it always entails a degree of disruption. In our case, the decision to bypass the rule by managing Deviations in the E/View will generate confusion later in the process, when engineers will try to propagate a change across diﬀerent M/Views. Since the M/Views are modified independently, there would not be a common starting point for the incremental change; any further modifications would have to be done therefore on an ad hoc basis, which is time-consuming, complicated and error-prone. These were precisely the type of drawbacks that the software was introduced to eliminate. Eluding the software’s sequential logic therefore implies duplication of work and a higher risk that subsequent changes will be implemented incorrectly: For example, a request arrived to ERS, asking to make a further change to an Order that had been made ‘Batch-2 eﬀective’ for the under-bodies; the problem

The influence of artefacts and distributed agencies on routines’ dynamics 203 was that, according to the request, the further change had to be made ‘Batch-1 eﬀective’. Theoretically this would be impossible, because the other change was Batch-2 eﬀective and one cannot backdate eﬀectivities, in theory. The reason that the first change had been made Batch-2 eﬀective in the first place, was that there were complex interrelationships of parts and eﬀectivities; and it was decided that the easiest thing was to make the whole thing Batch-2 eﬀective. But then all the aﬀected items on the change became Batch-2 eﬀective, and now they are prevented from making that change Batch-1 eﬀective. (Interview/MC)

These problems are generated by the complex interrelationships between parts (structural complexity) and between ‘eﬀectivities’ of diﬀerent kinds (time and process complexity). ‘You get a huge merry-go-round that you have to untangle’ (ibid.). The control action exerted by software represents one way to handle such complexity; rules and SOPs act as stabilizing factors in the context of the unstable and disordered development process. Software rules attempt to discipline the process by imposing a sequential and ordered logic; they can help to ensure that the work undertaken on one product part is consistent and concurrent with work undertaken on related parts elsewhere in the organization: A classic example of a rule is that you cannot release revision D of a part until after you have released revision C. One of the engineers gave us a Deviation saying: ‘I want revision F of this part in the BoM’. But what is released today is just revision C. He hasn’t released D, he hasn’t released E or F, but he wants revision F in the BoM. He has already sent information to the supplier, and the supplier is going to produce the part according to revision F specifications. The problem is that that part is related to other parts, and the other parts have not been released to a revision that matches that part’s revision, so you end up with a very long chain of interrelationships that you have to resolve. (Interview/MC)

PDM is conceived to introduce control and to clarify structural and process-time relationships. Nevertheless, there are circumstances where the action exerted by PDM is perceived as reducing flexibility: ‘PDM is introduced to control and discipline, but sometimes it gets in the way’ (Interview/MC). A production-oriented system, PDM is designed to support data control and validation which is mandatory in the downstream Production and Manufacturing environments, as argued by a Manufacturing manager: We need these BOMs; we need to schedule these BOMs into vehicles; and we need to have a Material Required Date for each vehicle [. . .] With that kind of [unreliable] scheduling, you just can’t do just-in-time. So you schedule one MRD date, and that is actually the MRD date for vehicle 1. Everything else, [and] there will be no time control. And of course we are talking about thousands of parts. And one vehicle every two weeks. You need a system to control that and you need people to operate that system. (Interview/JK)

204

Organizational routines and organizational change and innovation

The PDM Project Manager shares a similar view: ‘[A] lack of procedures breeds low quality due to [the] lack of data integrity, history and process repeatability – [this is] not sustainable. PDM will provide an environment to support [controlled] procedures.’ (Interview/DA) The Management and Manufacturing views are reflected in and supported by the software incremental change rule. However, the way they see the rule upstream in Engineering is quite diﬀerent: That is a very good rule. But the problem is that we live in a very anarchic world, in development, there is very little time and there are very few resources. So it may be the case that even if the guy has not released his revision F of the part, he still wants to specify that he wants revision F in the vehicle. (Interview/MC)

4.3

Mutual Adaptation between Formal and Actual Process

We have witnessed the unravelling of tensions between actual practices, on one hand, and the software procedure, on the other, which embodies assumptions and rules aimed at ordering and disciplining those practices. In this context, it appeared clear that the software’s emphasis on control was often perceived as excessive for those functions that rely on flexibility in both process and product structures. This can result in practitioners working around the rule to attempt to restore such flexibility. In other words, where the philosophy around which the software system is built and configured clashes with upstream engineering views of development, scope is created for deviations from the rule: A good example is this freeze. There will be a freeze. This is a rule, but then, in practice, this becomes only an ideal date by which it would be nice if everything would be out for these guys [read: programme management]. [. . .] It is a schizophrenic world, everyone will tell you that there are rules, and that, of course, we must work according to the rules, but then they will immediately break those rules. (Interview/MC)

Workarounds can favour the process of adaptation by reducing the dissonance between the worldviews of upstream (Industrial Design, Engineering) development functions, on one hand, and downstream (Production, Manufacturing) functions, as mediated by and embedded in software, on the other. We have also seen, however, that relaxing control by delaying the configuration freeze at the design stage is not without consequences. While excessive control can cause rigidity at the Engineering end, the lack of control becomes a liability in Production, where practitioners are faced with excessive change as well as having to work with late and unreliable data. While in fact at the Engineering end variation is highly desirable, in

The influence of artefacts and distributed agencies on routines’ dynamics 205

a Production environment control is mandatory, as one engineer convincingly put it: ‘For our prototype we want to be able to control the anarchy. In production you definitely want sequential incremental change’ (ibid.). Excessive deviation from the rule would work against the very purpose for which software-embedded rules were introduced in the first place, that is avoiding duplication of eﬀorts and facilitating change control, therefore providing Manufacturing with stable and reliable releases. One should: Come back to the basic question: what is the reason for the freeze? The reason for the freeze is that Manufacturing is incapable of coping with the huge volume of change; so we are going to freeze it. We say: this spec is what we are going to build it to. And there will be a minimal amount of change after that time. (ibid.)

This view, shared by Manufacturing and Programme Management, eventually prevails in this case. The fact that the rules and SOP are embedded in software helps them to endure against the other agencies’ views. Practitioners in Engineering were able to deviate from the course of the SOP but only to an extent, and for a limited period of time. Inscribing the SOP and rules in software has made them more visible and easier to enforce. In other words, once embedded in software, SOP and rules as statements have successfully managed to construct a world in which they can function. This world unravels as a result of the co-performation of competitive programs, in this case reflected by the Engineering vs. the Management and Manufacturing views of the world.

5.

DISCUSSION AND FRAMEWORK

In this chapter we have addressed three questions. First, how can we theorize about the interactions between SOPs and rules on one hand, and performances on the other? Second, what is the role played by artefacts in mediating these interactions? And third, what is the influence of distributed agencies, including heterogeneous organizational communities, in shaping the co-evolution between diﬀerent aspects of routines? Going back to and building on Callon and MacKenzie’s notion of performativity we can now shed new light on the co-evolution of formal rules and SOPs (the model) with routines in context (the actual process). 5.1

The Performativity of Routines

In our example, formal routines, rules and SOPs, embodying one or more of the multiple, abstract (ostensive) views of the freeze process, act as

206

Organizational routines and organizational change and innovation

models or guiding principles for the actual process (performances). SOPs are produced via a process of articulation, codification and standardization, through which actual processes are articulated or made explicit and ‘disentangled’ from the local context. Our ‘freeze’ procedure is one such ‘disembodied’ process. In order to disentangle the SOP from the actual process, a precise boundary must be drawn between the actions and knowledge inputs that are allowed by the SOP and those that are disallowed. What is left out by this framing process (the overflow) is a testimony to the eﬀortful work of codification and standardization that made the relationships among people, parts of the artefact and process, and parts of the organization explicit, thus giving the SOP a role as template for the development process. As a ‘disentangled object’, the SOP can provide a common reference point to coordinate heterogeneous knowledge and views across diﬀerent communities (that is Engineering and Production), therefore helping to coordinate and direct actions. Further, the processes of codification and standardization allowed for the ‘freeze routine’ to be disentangled and embedded in software, thus making it visible and available to all functions that could from then onwards use it to coordinate and synchronize their work. Framing has made the freeze process more visible and predictable, by making relationships (interactions, deliverables and deadlines) explicit and reducing ambiguity. The process of de-contextualization, through codification and the consequent inscription of the process in software, has transformed the freeze process into something which is easier to describe, visualize, share, transfer and reproduce (at least in principle) ‘anywhere and anytime’ across heterogeneous communities and multiple organizational locations. In other words, the process thus created was something more similar to a ‘standard’ (Bowker and Star, 1999), a commodity (cf. Callon, 1998) and could subsequently act as a reference ‘common model’ (D’Adderio, 2001), a single and relatively stable ‘interpretative frame’ of central process (March and Simon, 1958; Marengo, 1996; West and Iansiti, 2003). Our SOP was introduced with the specific intention to enforce and sustain global coordination and concurrency of work across heterogeneous development functions and teams. Once established, the SOP began to serve as a visible ‘frame of reference’ able to reduce cognitive complexity (Simon, 1947; Prietula and Augier, 2005) as well as constraining and guiding divergent intentions, views and actions across the organization. We have seen, however, that such a strong control mechanism, whose influence was further reinforced by inscribing the SOP in software, was perceived as excessively rigid by upstream functions where the need for change was greater. As argued in earlier work (D’Adderio, 2003), the process of codification and standardization in the production and reproduction of

The influence of artefacts and distributed agencies on routines’ dynamics 207

routines and SOPs is never neutral – as economic diﬀusion theory would have it – but is always performed according to one – or more – prevailing logics, views or rationales. It follows that our standard operating procedure is not merely a simplified version of the actual routine. The SOP in our example embodies a strong rationale supporting the views and objectives that belong to specific agencies, namely the Engineering Administrators, Programme and Project Management and Production and Manufacturing philosophy, at the expense of those belonging to upstream development functions (that is Industrial Design, Design Engineering). This emphasis highlights potential sources of conflict amongst diﬀerent agencies and their competing performative programmes.20 One calculative agency can impose its own calculations and rules, consequently forcing other agencies to engage in its own calculations, as in Callon’s example of the notorious chess player: ‘it is almost as if Kasparow [. . .] had to start calculating his moves not by playing like Kasparow but by imagining himself in the computer’s position, that is to say by borrowing from its algorithms and calculation rules’ (Callon, 1998: 46). The choice of a tool such as our software package can shift the balance in the competition between calculative agencies in favour of one agency or another; the probability of gain is in fact ‘on the side of the agency [. . .] whose tools enable it to perform, to make visible and to take into account the greatest number of relations and entities’ (Callon, 1998: 44–5). It follows that ‘Imposing the rules of the game can be done by imposing the tools in which these rules are incorporated’ (ibid.: 46). In our example, the inflexibility of the software-embedded SOP whose philosophy clashed with the goals, views and resources that belonged to other functions, generated scope for overflowing, manifested as a deviation from the SOP. The deviation took the form of a workaround, which temporarily restored flexibility while preventing the freeze process from grinding to a halt. The software procedure, thus, was only partially and temporarily bypassed. This was partly due to the realization that excessive variation from its course would have caused a loss of coordination and synchronization among development functions and across these and the rest of the organization – objectives sought after by both IT project management and PD programme management. The potential benefits of the software SOP in this case would not have been realized: global coordination would have been impossible without the framing, which provided a common reference point about the states of product and process, recording possible actions, their timing and their expected outcomes. Analogously to Callon’s ‘space for calculability’, our SOP acted as a standard, providing a common language, which enabled it to ‘reduce heterogeneity’ and ‘construct equivalence’ (Callon, 1998: 22).

208

Organizational routines and organizational change and innovation

Another reason why the SOP was not completely bypassed was the fact that, once embedded in software, controls become more invasive, pervasive and diﬃcult to detect, modify or altogether remove. Modifying the code underlying the software, as argued earlier, would have required resources such as time and complex programming skills that were simply not available. In addition, discarding the software-embedded rule would also have made the entire software philosophy void, invalidating the rationale underlying the expensive and resource-consuming software implementation programme. A process of reframing was thus instigated whereby high-level software-embedded rules were (mostly) held in place and abided by. This analysis helps to clarify the crucial but subtle dynamics that have so far been overlooked within the routines debate by revealing the deeper interactions between aspects of routines, artefacts and distributed agencies. Performativity theory has allowed us to break down the fictitious boundaries between ‘animated agencies’, on one hand, and ‘inanimate objects’ or structures, on the other, thus helping us to bring formal routines, rules and artefacts to life.21 We have thus been able to uncover the socio-technical processes by which rules and procedures emerge (that is how abstract views by diﬀerent agencies come to be embedded in rules, SOPs and other artefacts), how they influence performances of the freeze routine and how they are influenced in return (in other words, how they are performed). We have thus shown that, while of course people are free to interpret rules, and rules are never totally binding as they imply irreducible margins of interpretation, formal rules do perform a function. As cogently argued by our engineer, rules ‘Are there for a reason’. Artefact-embedded rules and procedures constrain interpretation and shape subsequent action, and are often shaped in return. Just as economic theory in MacKenzie’s and Callon’s examples is a constituent part of the market, our rules and SOPs are performative with respect to the actual ‘freeze’ process. As a final step, we can now relate each of MacKenzie’s (2006b) performativity categories to our case evidence. We have shown that aspects of the SOP and related rules were actually used by engineers in the freeze process (the deadline, the configuration handover procedure) as in ‘generic performativity’. Further, aspects of our SOP and rules had an eﬀect on the freeze process by instigating the emergence of a centralized, mostly visible, concurrent process (‘eﬀective performativity’). As a consequence, the practical use of an aspect of a rule or SOP made the freeze process more like its model (with the uploading of the ‘frozen’ configuration on the computer system) as in ‘Barnesian’ performativity; in these cases the SOP was initially able to provide guidance and discipline the process, instigating convergence between the formal procedure and the actual performance; this continued up to the point where the SOP was perceived as being too

The influence of artefacts and distributed agencies on routines’ dynamics 209

rigid. At that point the actual process started to diverge significantly from the model; in other words, the practical use of an aspect of our rule and SOP actually made our freeze process less like its depiction by the SOP (the creation of the workaround), and its influence was therefore ‘counterperformative’. Finally, through the addition of a formal deviation document to the (unlawfully) revised part, the workaround was finally legitimized and embedded in the formal computer-managed SOP. Divergence was contained in this case by the prevailing management view that local drift would have oﬀset the benefits of global coordination aﬀorded by the SOP. The inscription of part of the informal workaround in software steered the process again towards convergence, but this time it was the SOP that was modified to resemble the actual process. This made the formal procedure more flexible than it had been initially. Our evidence has thus captured the full cycle of adaptation between SOPs/rules and performances, and the role of artefacts as intermediaries in these interactions. In so doing we have shown that formal, artefactual representations of routines (rules and SOPs) do not solely ‘guide’ performances, as often argued in literature (cf. Blau, 1955), but they are performed through iterative cycles of framing, overflowing and reframing. 5.2

Prescription, Description, Performativity, and the Role of Artefacts

We can now return to our earlier distinction between the mechanistic and interpretive schools of thought in characterizing the influence of SOPs and rules. Our evidence demonstrates that each school has placed itself at one of the extreme ends of MacKenzie’s performativity chart (Figure 9.4). At one extreme (represented by the mechanistic school of thought) is the view that procedures and rules completely prescribe actions. In interpreting the role of SOPs and rules as deterministic and equating actors to rule-following automata, this view corresponds to the ‘framing’ side of our performativity spectrum (see Figure 9.5). While we accept that SOPs and rules frame actions and viewpoints to an extent, we have seen that full prescription is a rare and extreme outcome which entails the absence of adaptation, no resistance and automatic reproduction. At the other extreme is the agency-centred school which sees SOPs and rules as merely descriptive: a simplified copy of the actual process which they (often inadequately) attempt to mirror. This view, centred on overflowing, highlights the interpretive role of human agencies which, in enacting rules, are able to modify or completely dismiss them (rules in this case are counter- or non-performative) (Figure 9.5). While this is possible, this view does not account for the argument that, by incorporating beliefs into material devices, algorithms, procedures and routines a model can have

Organizational routines and organizational change and innovation

210

Non-performativity (demise/disuse) Generic performativity SOP, rule is used in the process

weak

Effective performativity SOP, rule has an effect on the process Barnesian performativity SOP, rule makes process more like its depiction Counter-performativity SOP, rule makes process less like its depiction

strong

Prescription (algorithmic sequence) Source:

Adapted from MacKenzie (2006b).

Figure 9.4

Degrees of performativity

an eﬀect ‘even if those who use them are sceptical of the model’s virtues, unaware of its details, or even ignorant of its very existence’ (MacKenzie, 2006a: 19). While, thus, formal procedures and rules can always – in theory – be worked around and dismissed, in practice they often play a role. Especially when embedded in artefacts such as software, and/or entangled in thick organizational interrelationships, they become visible, pervasive, diﬃcult to change or avoid, easier to enforce. While possible in theory, mechanistic prescription and full interpretive flexibility are in practice two extreme outcomes; to the extent that a rule is entangled in a web of tools and organizational relationships, some level of performativity is at play (Figures 9.4 and 9.5). Artefact-embedded SOPs and rules thus don’t simply describe, don’t often prescribe, mostly they are performed. In other words, they are ‘engines, not cameras’ (MacKenzie, 2006a). 5.3

Performativity Struggles and the Influence of Organizational Communities

To conclude our discussion we want to characterize the role of distributed agencies in routines evolution. A performative view allows us to characterize the emergence of rules and routines as the outcome of

The influence of artefacts and distributed agencies on routines’ dynamics 211

(Re-)Framing (convergence)

Rules, SOPs

Performances

Overflowing (divergence) Figure 9.5

Theoretical framework

performativity struggles among competing ‘agencements’ that aim at constructing the (industrial freeze) process in diﬀerent manners. We have argued that the more successful performative programmes are those that manage to enrol an array of materials and tools to create a world in which they can function. In our case, Management and Manufacturing manage to enrol the software and thus to impose (at least partially) the culture and priorities that belong to their occupational communities, their idiosyncratic ‘languages’ and worldviews (Galison, 1999; D’Adderio, 2001).22 They are thus able to do the classification and ordering of data and process and decide what is and what is not important, what is allowed or disallowed. Once embedded in software, the SOP has become a very powerful statement: it is visible to all functions, spans across all relevant organizational boundaries and communities, makes management intent clear and unequivocal and provides a means of comparing legitimate with illegitimate actions and viewpoints. The statement has managed to put its world into motion. Just like the uploading of the virtual product in software in D’Adderio (2001) and the tracing of price variation curves in Preda (2007), our software-embedded SOPs and rules, as inscriptions, impose a principle of reality; they constitute an obligatory point of passage, a perfectly material reality to take into account [. . .] They are articulated to sociotechnical agencements that produce the traces that they use to inscribe the world

212

Organizational routines and organizational change and innovation

in which they are participants and on which they will, in turn, make possible to act. (Callon, 2007, pp. 39–40)

While our evidence could thus be interpreted as a mere clash between conflicting interests, there is more to it. On one hand, organizational groups hold diﬀerent views of the freeze routine. They also have diﬀerent incentives and success criteria for the routine: for upstream engineers the criterion for success is the production of the best (most innovative, robust) design; for Management, success is about getting the product out of the door on time; for Manufacturing it is about having a stable, well-tested product configuration; for PDM project management it is about getting every function to use the software to achieve greater control and speed up the development process; for engineering administration, success is about the seamless integration of upstream design and downstream manufacturing data. If, however, on the other hand, we reduced this complex reality to a mere clash between conflicting interests, we would fall short of capturing what is truly happening. To paraphrase Callon, the various organizational actors attempt to construct the world (or socio-technical agencement) they believe to resemble their own assumptions, views and aims. Confrontation therefore takes place not simply between diﬀerent agencies but diﬀerent worlds that are struggling to exist, one at the expense of the other (Callon, 2007). The result of these struggles is that often none of the actors are able to take their programme to its conclusion, since no one function is able to frame the (engineering freeze) process exclusively. Each has to compromise and at least partially accept the others’ programme, meaning that only portions of their world are realized. In our example, Programme and IT management and manufacturing partially succeeded in disciplining actions according to the software freeze rule – therefore enforcing the deadline – and yet engineering retained some of their discretion to make substantial and late alterations. Thus the world that ended up existing was a compromise, a patchwork containing elements from competing worlds. Each of the agencies thus managed to exert some influence over the overall process: in the end, consistently with Callon’s argument, the losers in such a complex situation are only the worlds that are excluded – or exclude themselves – a priori. By deciding to retain their legacy software, and thus not to engage with the organization-wide PDM implementation, for example, Industrial Designers find that it has become increasingly hard for them to bring their inputs, views and requirements to bear upon the rest of the development process.

The influence of artefacts and distributed agencies on routines’ dynamics 213

6.

CONCLUSIONS

Shifting the emphasis from routines as undiﬀerentiated monolithic ‘objects’ to routines as generative – and continuously emerging – systems characterized by internal structures and dynamics provides promising new ground for exploring some of the most relevant but as yet under-researched questions about the nature and dynamics of routines. These include the dynamics of interaction between aspects of routines and the role of artefacts and agencies in shaping these dynamics. This work has made a contribution towards filling these important gaps in our understanding of routines dynamics. The combined novel empirical focus on artefacts and theoretical framework based on new developments in Economic Sociology and Sociology of Finance have provided important new insights into the mutual adaptation between aspects of routines and the role of artefacts and other agencies as intermediaries. This approach has allowed us to move beyond fully prescriptive, normative approaches, on one hand, and simply descriptive, interpretive approaches, on the other. In accordance with the most recent advances in economic sociology we have thus shown that prescription is an extreme type of performation, and the interpretive flexibility of social actors is not absolute, as worldviews and theories – especially when embedded in artefacts – tend to play a role. Specifically, our framework has provided three main contributions to the routines debate. First, we were able to capture the micro dynamics of interaction between diﬀerent aspects of routines, namely, artefactual representations of routines (SOPs, formal rules) and actual performances. Drawing from recent theoretical developments within the field of STS, we have theorized routine evolution and adaptation as the emergent result of iterative cycles of framing (selective retention), overflowing (variation) and reframing (selective retention) by which SOPs and rules are performed.23 This framework helps us understand how a routine’s stable pattern emerges out of the mutual cycles of adjustment (convergence and divergence) among these elements and the competitive arrangements in which it is stabilized. Such stability, however, is constantly put into question. So, while from a distance the routine might look the same, in reality it is continually changing, ‘tuned and retuned in the struggles in fields of agency that the performative idiom thematises’ (Pickering, 1994: 415). As our evidence thus suggests, such struggles among performative programmes underpin the dynamics of convergence and divergence between routines’ constituent parts or aspects (in other words the direction of their interactions, that is performativity vs. counter-performativity), the intensity of these interactions (that is weak vs. strong performativity) and, ultimately, the persistence of the routinized

214

Organizational routines and organizational change and innovation

pattern or deviation from its course (depending upon which programme – or configuration of programmes – manages to prevail). Second, we have characterized how artefacts as intermediaries shape the interactions between diﬀerent sides of routines. In contrast with much extant literature that sees SOPs and rules as either imperfect process representations that can be easily dismissed/disused or as prescriptions that are compulsively and automatically reproduced, we have shown that – once they are embedded in a web of technological artefacts and organizational relationships – some kind of adaptation occurs. Therefore while rules do not suggest their own correct application, and in certain cases actors are able to enact their own interpretation of the rule, on the other hand, rules (especially when embedded in artefacts such as software) do have an influence. In other words, while SOPs and rules can be – in extreme cases – fully descriptive (a passive, fixed representation of the actual process) or fully prescriptive (univocally ordering and structuring the process), mostly they are performed. Here the distinction between performativity and prescription – while a matter of degree – becomes relevant: performation refers to uncertain situations where there is dynamic adaptation, while prescription refers to automatic reproduction and pure repetition. The latter represents an extreme case where the socio-technical agencements and the worlds corresponding to its models have been realized, and there is therefore no more recalcitrance, roles are performed automatically, and events are recurrent. When this type of adaptation occurs, the performation comes to resemble a prescription. This situation, however, is quite diﬀerent from our case, and indeed from most firms that use and produce complex technologies, whose operations are distributed across diverse organizations and heterogeneous work communities, and who are faced with uncertain environments and subjected to rapidly changing innovation and regulatory regimes.24 These are cases where counter-performativity prevails and existing agencements have to be rearranged or even profoundly transformed in order to become successful. This takes us to the third issue dealt with in this chapter, that is the role and influence of heterogeneous, distributed agencies such as occupational communities and CoPs (Lave and Wenger, 1991) on routine evolution and adaptation. In analysing how abstract views of routines become embedded in artefacts, we were able to account for the fundamental role of distributed, and often conflicting, agencies in shaping routines.25 This entailed an important shift of emphasis from the existing paradigm where routines constitute the structure and individuals the agency to socio-technical agencements that involve at the same time people and artefacts, material and non-material elements. We have shown how, in the struggle between

The influence of artefacts and distributed agencies on routines’ dynamics 215

competitive performative programmes, some agencies are able to inscribe their own worldviews in artefacts. These agencies are the most likely to succeed in exerting their own influence: enrolling artefacts tends to create stronger agencements that are more stable, are tightly interconnected into the web of organizational relationships and therefore are more diﬃcult to oppose. Rule-following is indeed an important form of distributed cognition. In our quest to unravel routines’ internal dynamics we have set our focus on artefacts – and, specifically, on SOPs and rules – as starting points for our analysis. Artefacts such as written rules and procedures – especially when embedded in software as in our case – have provided vantage points to observe the ostensive (abstract) aspects of routines with respect to which they can serve as indicators or ‘proxies’ (Pentland and Feldman, 2005b). We have thus embraced a pragmatic view of meanings and understandings which sees these as not simply residing ‘in people’s heads’ but as distributed across a thick organizational web including people, everyday artefacts, tools and procedures. For this reason, neglecting to include tools and artefacts in the study of routines dynamics can only provide at best a partial picture. So where existing literature has rightly emphasized the individual agent’s ability to ‘turn exceptions into rules’ (Feldman and Pentland, 2003: 110), we have shown that they can do so only to the extent that they are able to construct a successful agencement, which in turn often entails enrolling tools and technological artefacts. Indeed, as MacKenzie has shown with his example of the Black, Scholes and Merton formula – while the alignment of beliefs, views and intentions can work for a while, these tend to provide temporary arrangements unless they are able to create a world in which they can function (2003). This is done by enrolling a ‘principle of reality’, which is often achieved through the involvement of material objects and artefacts.26 We can therefore argue that (stable) SOPs and rules emerge not so much as the result of pure beliefs alignment but as the emergent outcome of competing agencements, some of which are more and some less able to enrol materials and therefore are more or less successful. In this sense, the fact that a procedure or rule ‘works’ is the result – and not the premise – of successful performation, a formula that – over time – has been able to create the world in which it can function and therefore now encounters little or no resistance. Performativity struggles between competing agencements lead to their mutual adjustment involving the (temporary) predominance of a strong programme, or the emergence of a new programme from the coexistence/ assemblage of diﬀerent ones. The resulting stability is indeed similar to Nelson and Winter’s notion of ‘truce’ (1982) – in its dynamic inception as a continuously challenged and emergent achievement – but here we can

216

Organizational routines and organizational change and innovation

see clearly what the forces at play are that are responsible for stabilizing or destabilizing the routine. By means of a new framework that builds on Performativity Theory’s set of notions and constructs and a fine-grained analysis based on ethnographical data, we were thus able to achieve valuable progress towards an improved characterization of routines dynamics and the fundamental influence of artefacts and agencies on their evolution.

ACKNOWLEDGEMENTS I would like to acknowledge the financial support of the ESRC grant for ‘The biography and evolution of standardised software packages’ (RES-000-23-0466) as well as my personal Innovation Fellowship grant entitled ‘Making innovation dependable: validating complex technologies across the network organisation’, sponsored by the Advanced Institute of Management Research (AIM) (RES-331-27-0014) which supported the final stages.

NOTES * 1.

2. 3.

4.

5. 6.

Reprinted from Research Policy, 37(5), Luciana D’Adderio, ‘The performativity of routines: theorising the influence of artefacts and distributed agencies on routine dynamics’, pp. 769–89, Copyright 2008, with permission from Elsevier. In this chapter we focus on formal rules and procedures thus neglecting other categories of ‘stable systemic traits’ (Cohen et al., 1996) such as ‘rules of thumb’ and ‘heuristics’. We also assume that, while explicit rules and procedures are not the same as formal statements they can aﬀord the same analytical treatment. See also D’Adderio (2003). The notion of distributed cognition and distributed agency was introduced by Ed Hutchins in his work on navigation (1995). He argues that to understand cognition ‘in the wild’ one must go beyond the analysis of the individual ‘bounded by the skin’ as cognition frequently involves multiple collaborating human beings interacting with artefacts: ‘local functional systems composed of a person in interaction with a tool have cognitive properties that are radically diﬀerent from the cognitive properties of the person alone’ and a ‘group performing [a] cognitive task may have cognitive properties that diﬀer from the cognitive properties of any individual’ (Hutchins, 1995: xvi and 176). Our framework builds on Cohen et al.’s (1996) distinction between ‘routines-asrepresentations’ and ‘routines-as-expressions’. At the same time, however, it is compatible with Feldman and Pentland’s distinction between ‘ostensive’ and ‘performative’ (2003) with artefactual representations characterized as ‘proxies of the ostensive’. The term ‘performative’ as used here is borrowed from Performativity Theory (Barnes, 1982; Callon, 1998; MacKenzie, 2003). Cohen et al.’s (1996) distinction between ‘representations’ and ‘expressions’ originates from the biological distinction between genotype and phenotype whereby a genome stores the pattern-guidance needed to reproduce its phenotypic expression, and the modifications of the genetic representation are the source of evolutionary variation. This

The influence of artefacts and distributed agencies on routines’ dynamics 217

7.

8.

9.

10.

11. 12.

13. 14.

15.

distinction – akin to Organization Theory’s traditional quest to account for the roles of both formal and informal organizational features (Simon, 1947) – provides the basis for a theoretical characterization of the diﬀerent but complementary roles of SOPs and performances: while ‘SOPs-as-representations’ consist of formalized statements of what actions should occur, the actions occurring as routines are expressed in context. STS scholars’ exploration of the distinction between formal procedures and rules and informal practices is relevant here. Authors have distinguished between ‘mental plans’ and ‘situated action’ (Suchman, 1987; Hutchins, 1991), ‘representations of work’ and ‘practical action’ (Suchman, 1983), ‘modus operandi’ and ‘opus operatum’ (Bourdieu, 1977), ‘espoused’ and ‘actual’ practice (Orr, 1990), ‘time-objective’ and ‘time-in-process’, ‘object-world’ and ‘process-world’ (Bucciarelli, 1988), ‘rules-as-represented’ and ‘rulesas-guides-in-practice’ (Orr, 1990 in Tsoukas, 1996), ‘canonical’ and ‘non-canonical’ practice (Brown and Duguid, 1991), the ‘map’ and the ‘terrain’ (Berg, 1997). Continuous ad-hoc ‘articulation work’ (Suchman, 1983, 1987) of actors is thus constantly required to bridge the ‘phronetic gap’ (Spender, 1989; Taylor, 1993) between the formula (or rule) and its enactment, the representation and the actual process. While this literature identifies the tensions between the two categories of elements, it does not address their mutual adaptation. Contributions inspired by this literature so far have failed to emphasize that the interactions between ‘expression’ and ‘representations’ are far from straightforward as they involve the radical reconfiguration of both the form and the content of routines (cf. D’Adderio, 2003). The outcome of the complex, socio-technical process of translation from artefactual representation or code (Hutchins and Hazelhurst, 1991), to a new expression-type routine, for example, may contain parts of representation-type routines as the boundaries separating them are often undefined. The fundamental diﬀerence between the ‘social constructivist’ and the ‘ethnomethodological’ approaches is that the former acknowledges the role played by constraints – from the psychological to the sociological – in the process of attributing meanings to rules (Bloor, 1997). While falling short of including technology among the constraints, constructivists nonetheless accept that these allow foreclosure in practice. Mambrey and Robinson (1995), for example, have shown how, in a highly hierarchical and rational organisation, existing rigid procedures are often dismissed in favour of intricate, informal processes (see also D’Adderio, 2004); and Wynne (1988) showed how operators of large technical systems often deviate from formal, rule-binding operating practices to deal with complex interdependencies, unexpected circumstances and local conditions. Similarly, Gasser (1986) and Orr (1996, 1998) have discussed how a computer’s formal representation of the work clashes continuously with the actual contingent and complex logic of human work. Since the formal tool embodies an impoverished version of work, humans working with the tool need to repair the tool’s functioning whenever it is used in practice (Collins, 1990). The term ‘script’ is the actor–network theory (ANT) equivalent of the notion of rule in sociological inquiry (Preda, 2000). Latour’s notions of ‘immutable mobiles’ and ‘obligatory points of passage’ (1987) play an important role in this context, though one that has been neglected so far in the organizational literature that has mainly focused on artefacts as flexible intermediaries or ‘boundary objects’ (Bechky, 2003). In contrast, D’Adderio (2001) has shown how coordination across diverse organizational communities requires a mix of ‘boundary objects’ and ‘standardizing devices’. On the influence of information systems on structuring knowledge and work see D’Adderio (2003). As information systems grow both in scope and complexity, fewer user organizations possess the necessary competences and skills required to adapt the software to their practices; as a result they often come to rely on the expensive expertise of external consultants or software producers to customize the code. A socio-technical agencement is the assemblage of heterogeneous elements that is required for the world contained in the statement to be actualized: ‘A formula that

218

16. 17. 18.

19. 20.

21. 22. 23.

24. 25. 26.

Organizational routines and organizational change and innovation progressively discovers its world and a world that is put into motion by the formula describing it’ (MacKenzie, 2003). Product Data Manager (PDM) is a leading state-of-the-art, commercial oﬀ-the-shelf (COTS) enterprise software application for integrated design and manufacturing. On the use of narratives as instruments for empirical analysis see Orr (1990). The philosophy behind this reflects established industry trends that see increasing the speed of the development cycle while increasing the eﬃciency of data integration across development functions as major sources of advantage in the fiercely competitive automotive market. Computer-Supported Cooperative Work. Authors have shown how diﬀerent worldviews by diﬀerent organizations/agencies are reflected in the multiple ostensive aspects of routines (Adler et al., 1999; Feldman, 2003, see also Pentland and Feldman, 2005a). Here we add to this work by showing how at times some agencies are able to impose their worldview by enrolling tools, artefacts and procedures, thus creating stronger agencements that, at least temporarily, help their views and goals to prevail at the expense of others. On the issue of artefacts and agency see also Latour (2005). For a discussion of the literature on organizational communities see also Brown and Duguid (1996) and Cohendet and Llerena (2003). This view is aligned with the evolutionary framework where endogenous change emerges from cycles of interactions between performances or expressions (variation) and coded artefacts or representations such as SOPs and rules (selective retention) (Cohen et al., 1996). This is compatible with but diﬀerent from Feldman and Pentland’s framework where performances provide variations which are selectively retained in the ostensive aspect of the routine (see Becker et al., 2005 for a discussion). In these cases, ‘it is more diﬃcult for performations to produce regularities and repetition as they are constantly faced with unexpected events that they sometimes absorb, but only sometimes, for a while’ (Callon, 2007: 61–2, emphasis added). The topic of routines and governance has been largely neglected in the routines literature since Nelson and Winter (1982) and Coriat and Dosi (1994)’s pioneering contributions. Feldman’s (2000) example of the failure to implement a new routine for university housing can be given in this light a complementary reading as a failure to enrol people and artefacts to create a successful agencement that brings a procedure to life and, most importantly, keeps it operative.

REFERENCES Adler, P.S., B. Goldoftas and D.I. Levine (1999), ‘Flexibility versus eﬃciency? A case study of model changeovers in the Toyota Production System’, Organization Science, 10(1), 43–68. Akrich, M. (1992), ‘The de-scription of technical objects’, in W.E. Bijker and J. Law (eds), Shaping Technology/Building Society – Studies in Sociotechnical Change, Cambridge, MA, USA: MIT Press, pp. 205–24. Barley, S. (1986), ‘Technology as an occasion for structuring: evidence from observation of CT scanners and the social order of radiology departments’, Administrative Science Quarterly, 31, 78–108. Barnes, B. (1982), T.S. Kuhn and Social Science, London and Basingstoke, UK: Macmillan. Bechky, B.A. (2003), ‘Sharing meaning across occupational communities: the transformation of understanding on a production floor’, Organization Science, 14(3), 312–30.

The influence of artefacts and distributed agencies on routines’ dynamics 219 Becker, M.C., N. Lazaric, R.R. Nelson and S.G. Winter (2005), ‘Applying organizational routines in understanding organizational change’, Industrial and Corporate Change, 14(5), 775–91. Berg, M. (1997), ‘Of forms, containers, and the electronic medical record: some tools for a sociology of the formal’, Science, Technology and Human Values, 22(4), 403–33. Berg, M. (1998), ‘The politics of technology: on bringing social theory into technological design’, Science, Technology and Human Values, 23, 456–90. Blau, P. (1955), The Dynamics of Bureaucracy, Chicago, IL, USA: University of Chicago Press. Bloor, David (1973), ‘Wittgenstein and Mannheim on the sociology of mathematics’, Studies in the History and Philosophy of Science, 4, 173–91. Bloor, David (1997), Wittgenstein, Rules and Institutions, London, UK: Routledge. Bourdieu, P. (1977), Outline of a Theory of Practice, Cambridge, UK: Cambridge University Press. Bourdieu, P. (1990), The Logic of Practice, trans. R. Nice, Cambridge, MA, USA: Polity Press. Bowker, G.C. and S.L. Star (1999), Sorting Things Out: Classification and its Consequences, Cambridge, MA, USA: MIT Press. Brown J.S. and P. Duguid (1991), ‘Organizational learning and communities of practice’, Organization Science, 2(1), 40–57. Brown, J.S. and P. Duguid (1996), ‘Learning and communities-of-practice: toward a unified view of working, learning, and innovation’, in M.D. Cohen and L.S. Sproull (eds), Organizational Learning, London, UK: Sage. Bucciarelli, L.L. (1988), ‘Engineering design process’, in F.A. Dubinskas (ed.), Making Time: Ethnographies of High-Technology Organisations, Philadelphia, USA: Temple University Press, Ch.3, pp. 92–122. Callon, M. (1998), ‘An essay on framing and overflowing: economic externalities revisited by sociology’, in M. Callon (ed.), The Laws of the Markets, London, UK: Blackwell, pp. 244–69. Callon, M. (1999), ‘Actor–network theory – the market test’, in J. Law and J. Hassard (eds), Actor Network Theory and After, Oxford, UK: Blackwell, pp. 181–95. Callon, M. (2007), ‘What does it mean to say that economics is performative?’, in D. MacKenzie, F. Muniesa and L. Siu (eds), Do Economists Make Markets? On the Performativity of Economics, Princeton, NJ, USA: Princeton University Press. Cassell, P. (ed.) (1993), The Giddens Reader, Stanford, CA, USA: Stanford University Press. Cohen, M.D., R. Burkhart, G. Dosi, M. Egidi, L. Marengo, M. Warglien and S. Winter (1996), ‘Routines and other recurring patterns of organisations: contemporary research issues’, IIASA Working Paper, March. Cohendet, P. and P. Llerena (2003), ‘Routines and incentives: the role of communities in the firm’, Industrial and Corporate Change, 12, 271–97. Collins, H.M. (1990), Artificial Experts. Social Knowledge and Intelligent Machines, Cambridge, MA, USA: MIT Press. Coriat, B. and G. Dosi (1994), ‘Learning how to govern and learning how to solve problems: on the co-evolution of competences, conflicts and organisational routines’, prepared for the Prince Bertil Symposium, Stockholm School of Economics, June.

220

Organizational routines and organizational change and innovation

Cyert, R. and J. March (1963), A Behavioural Theory of the Firm, 2nd edn, Cambridge, MA, USA: Blackwell. D’Adderio, L. (2001), ‘Crafting the virtual prototype: how firms integrate knowledge and capabilities across organisational boundaries’, Research Policy, 30(9), 1409–24. D’Adderio, L. (2003), ‘Configuring software, reconfiguring memories: the influence of integrated systems on the reproduction of knowledge and routines’, Industrial and Corporate Change, 12(3), 321–50. D’Adderio, L. (2004), Inside the Virtual Product: How Organisations Create Knowledge Through Software, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Daft, R.L. and K.E. Weick (1984), ‘Toward a model of organizations as interpretation systems’, The Academy of Management Review, 9(2), 284–95. Eisenhardt, K.M. (1989), ‘Building theories from case study research’, Academy of Management Review, 14(4), 532–50. Feldman, M.S. (2000), ‘Organizational routines as a source of continuous change’, Organization Science, 11(6), 611–29. Feldman, M.S. (2003), ‘A performative perspective on stability and change in organizational routines’, Industrial and Corporate Change, 12(4), 727–52. Feldman, M.S. and B.T. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48, 94–118. Galison, P. (1999), ‘Trading zone: coordinating action and belief’, in M. Biagioli (ed.), The Science Studies Reader, New York, USA: Routledge, pp. 137–60. Garfinkel, H. (1967), Studies in Ethnomethodology, New York, USA: PrenticeHall. Gasser, L. (1986), ‘The integration of computing and routine work’, ACM Transactions on Oﬃce Information Systems, 4, 257–70. Giddens, A. (1993), New Rules of Sociological Method, Cambridge, UK: Polity Press. Glaser, B. and A. Strauss (1967), The Discovery of Grounded Theory: Strategies of Qualitative Research, London, UK: Weidenfeld and Nicholson. Grint, K. and S. Woolgar (1992), ‘Computers, guns, and roses: what’s social about being shot?’, Science, Technology and Human Values, 17(3), 366–80. Hatherly, D., D. Leung and D. MacKenzie (2007), ‘The finitist accountant’, University of Edinburgh, April. Hennion, A. (1993), La passion musicale, Paris, France: Métaillié. Holm, P. (2002), ‘Which way is up on Callon? A review of a review: Daniel Miller’s “Turning Callon the right way up”’, Norwegian College of Fishery Science, Tromsø, Norway. Hutchins, E. (1991), ‘Organizing work by adaptation’, Organization Science, 2(1), 14–39. Hutchins, E. (1995), Cognition in the Wild, Cambridge, MA, USA: MIT Press. Hutchins, E. and B. Hazelhurst (1991), ‘Learning in the cultural process’, in C.G. Langton et al. (eds), Artificial Life II: Studies in the Sciences of Complexity, Reading, MA, USA: Addison-Wesley. Latour, B. (1987), Science in Action, Cambridge MA, USA: Harvard University Press. Latour, B. (1992), ‘Where are the missing masses? The sociology of a few mundane artefacts’, in W.E. Bijker and J. Law (eds), Shaping Technology/Building Society:

The influence of artefacts and distributed agencies on routines’ dynamics 221 Studies in Sociotechnical Change, Cambridge, MA, USA: MIT Press, pp. 225–58. Latour, B. (2005), Reassembling the Social. An Introduction to Actor Network Theory, Oxford, UK: Oxford University Press. Lave, J. and E. Wenger (1991), Situated Learning: Legitimate Peripheral Participation, New York, USA: Cambridge University Press. Lazaric, N. and B. Denis (2001), ‘How and why routines change: some lessons from the articulation of knowledge with ISO 9002’, Economies et Sociétés, 6, 585–611. Lynch, M. (1992), ‘Extending Wittgenstein: the pivotal move from epistemology to the sociology of science’, in A. Pickering (ed.), Science as Practice and Culture, Chicago, IL, USA: University of Chicago Press, pp. 215–65. Lynch, M. (2001), ‘Ethnomethodology and the logic of practice’, in T.R. Schatzki, K. Knorr Cetina and E. von Savigny (eds), The Practice Turn in Contemporary Theory, London, UK: Routledge, pp. 131–48. MacKenzie, D. (2003), ‘An equation and its worlds: bricolage, exemplars, disunity and performativity in financial economics’, Social Studies of Science, 33(6), 831–68. MacKenzie, D. (2006a), An Engine, Not a Camera: How Financial Models Shape Markets, Cambridge, MA, USA: MIT Press. MacKenzie, D. (2006b), ‘Is economics performative? Option theory and the construction of derivatives markets’, Journal of the History of Economic Thought, 28(1), 29–55. Mambrey, P. and M. Robinson (1995), ‘Preparing a speech for the minister: notes towards understanding the role of artefacts in a flow of work’, Paper presented at ECSCW ’95. March, J. and H. Simon (1958), Organizations, New York, USA: Wiley. Marengo, L. (1996), ‘Structure, competencies and learning in an adaptive model of the firm’, in G. Dosi and F. Malerba (eds), Organization and Strategy in the Evolution of the Enterprise, London, UK: Macmillan. Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Belknap Press. Orlikowski, W.J. (1992), ‘The duality of technology: rethinking the concept of technology in organizations’, Organization Science, 3(3), 398–427. Orlikowski, W.J. (2000), ‘Using technology and constituting structures: a practice lens for studying technology in organizations’, Organization Science, 11(4), 404–28. Orlikowski, W.J. (2002), ‘Knowing in practice: enacting a collective capability in distributed organizing’, Organization Science, 13(3), 249–73. Orr, J.E. (1990), ‘Sharing knowledge, celebrating identity: war stories and community memory in a service culture’, in D.S. Middleton and D. Edwards (eds), Collective Remembering: Memory in Society, Beverley Hills, CA, USA: Sage Publications. Orr, J.E. (1996), Talking about Machines: An Ethnography of a Modern Job, Ithaca, NY, USA: ILR Press/Cornell University Press. Orr, J.E. (1998), ‘Images of work’, Science, Technology & Human Values, 23(4), 439–55. Pentland, B.T. and M.S. Feldman (2005a), ‘Organizational routines as a unit of analysis’, Industrial and Corporate Change, 14(5), 793–815. Pentland, B.T. and M.S. Feldman (2005b), ‘Designing routines: artifacts in support of generative systems’, presented in Sophia Antipolis, France, 21–22 January.

222

Organizational routines and organizational change and innovation

Pentland, B.T. and H.H. Rueter (1994), ‘Organizational routines as grammars of action’, Administrative Science Quarterly, 39, 484–510. Pickering, A. (1994), ‘After representation: science studies in the performative idiom’, Proceedings of the Biennial Meeting of the PSA, vol. 1994(2), 413–19. Pollock, N. (2005), ‘When is a work-around? Conflict and negotiation in computer systems development’, Science, Technology and Human Values, 30(4), 1–19. Pollock, N. and J. Cornford (2004), ‘ERP systems and the university as a “unique” organisation’, Information Technology and People, 17(1), 31–52. Preda, A. (2000), ‘Order with things? Humans, artifacts, and the sociological problem of rule-following’, Journal for the Theory of Social Behaviour, 30, 269–98. Preda, A. (2007), ‘Where do analysts come from? The case of financial chartism’ in M. Callon, Y. Millo and F. Muniesa (eds), Market Devices, Oxford, UK: Blackwell. Prietula, M. and M. Augier (2005), ‘Adventures in software archeology: seeking (ABTOF) theory in the code’, Working Paper, Goizueta Business School, Emory University, Atlanta, GA, USA. Sahlins, M. (1985), Islands of History, Chicago, IL, USA: University of Chicago Press. Simon, H. (1947), Administrative Behavior, New York, USA: Macmillan. Spender, J.C. (1989), Industry Recipes, Oxford, UK: Blackwell Suchman, L.A. (1983), ‘Oﬃce procedure as practical action: models of work and system design’, ACM Transactions on Oﬃce Information Systems, 1(4), 320–28. Suchman, L.A. (1987), Plans and Situated Action: The Problem of Human–Machine Communication, Cambridge, MA, USA: Cambridge University Press. Taylor, C. (1993), ‘To follow a rule’, in C. Calhoun, E. LiPuma and M. Postone (eds), Bourdieu: Critical Perspectives, Cambridge, UK: Polity Press, pp. 45–59. Taylor, F.W. (1911), The Principles of Scientific Management, New York, USA: Harper. Thomas, N. (1991), Entangled Objects, Exchange, Material Culture and Colonialism in the Pacific, Cambridge, MA, USA: Harvard University Press. Tsoukas, H. (1996), ‘The firm as a distributed knowledge system: a constructionist approach’, Strategic Management Journal, 17, 11–25. West, J. and M. Iansiti (2003), ‘Experience, experimentation, and the accumulation of knowledge: the evolution of R&D in the semiconductor industry’, Research Policy, 32, 809–25. Wittgenstein, L. (1967), Philosophical Investigations, Oxford, UK: Blackwell. Wynne, B. (1988), ‘Unruly technology: practical rules, impractical discourses and public understanding’, Social Studies of Science, 18, 147–67. Zollo, M. and S.G. Winter (2002), ‘Deliberate learning and the evolution of dynamic capabilities’, Organization Science, 13, 339–51.

10.

Innovation routines: exploring the role of procedures and stable behaviour patterns in innovation Markus C. Becker and Francesco Zirpoli

1.

INTRODUCTION

The capability to innovate, that is, to develop successful new products or processes, is important and highly treasured. One of its hallmarks, judging from the literature, is that the capability to innovate fluctuates considerably over the life time of a firm. Often, small firms are successful in developing new products particularly of the more radically innovative sort, while large firms seem to turn out new products that often represent mainly incremental innovations. Yet some companies have a track record of repeated radical innovation. Apple Computers is a famous example. In its long stream of successful new products, it counts the first personal computer, and recently, the iPod. Another example, just down the road in Silicon Valley, is IDEO, the design firm. It has generated innovations such as, for instance, the computer mouse. Both firms have a reputation for repeatedly developing successful new products, including a series of radical innovations. Apple has managed to bounce back successfully from phases of moderate success, most recently with the introduction of the iPod. IDEO has turned out a great number of new products with such continuity that it has been termed an ‘innovation factory’ (Hargadon and Sutton, 1997, 2000). The question raised by such examples is how some companies are able to generate innovations repeatedly, even including several radical innovations. In looking for an explanation of innovation outcomes, we are turning towards endogenous, rather than exogenous factors. Previous research has identified a series of factors that are supposed to have an impact on innovation outcomes (Cooper, 1990; Cooper and Kleinschmidt, 1995; Cooper, 1999). In this chapter, we scrutinize one of these factors in more depth than is often done: stable behaviour patterns in accomplishing innovation tasks. The underlying hypothesis is that repeated innovation outcomes – possibly even including a series of radical innovations – are caused by 223

224

Organizational routines and organizational change and innovation

some stable feature of the firm’s innovation activity. Such a perspective can be traced back to Schumpeter’s observation that innovation increasingly takes place in large firms where it becomes formalized and bureaucratized (Schumpeter, 1942). It is also consistent with Schumpeter’s interest in the ways in which firms go about running their business (Becker et al., 2006), a view that has been developed into the idea that to a large extent, a firm’s behaviour can be explained by its standard operating procedures (Cyert and March, 1963) or organizational routines (Nelson and Winter, 1982). From this vantage point, it would actually appear possible that firms become ‘innovation factories’ (Hargadon and Sutton, 1997, 2000) precisely because they follow ‘innovating routines’ (Pavitt, 2002). Such an idea starkly contrasts the notion that structuring and systematizing1 the new product development process increases the eﬃciency of the new product development process (in terms of lead time and cost), but is negatively correlated with innovation and creativity (at least of the more radical kind) (Burns and Stalker, 1961; Pierce and Delbecq, 1977; Aiken et al., 1980; Cardinal, 2001; Adler, 2007). A conclusion often drawn is that ‘the organization of non-routine tasks . . . central to innovation should be shielded from bureaucratic rationalization’ (Adler, 2007: 3; for examples of such research see Scott and Bruce, 1994; Ford and Gioia, 2000; West, 2002; De Jong and Kemp, 2003; Anderson and Gasteiger, 2007). Organizations should design units specialized in exploration and units specialized in exploitation, the latter heavily routinized and the former shielded from routinization and with more ‘organic’ structures in the spirit of Burns and Stalker (1961) (Tushman and O’Reilly, 1996; O’Reilly and Tushman, 2004; Gupta et al., 2006). There is scant empirical evidence supporting the hypothesis that routinization or formalization are negatively correlated with innovation, however (Ohly et al., 2006). Rather, evidence to the contrary is available, too, arguing that clearly specified work rules for the successful introduction of innovations in organizations are needed and that at least in some cases, routinizing the innovation process appears to be successful (Nelson and Winter, 1982; Benghozi, 1990; Griﬃn and Hauser, 1996; Wheelwright and Clark, 1992; Jelinek and Schoonhoven, 1993; Adler and Borys, 1996; Davila, 2000; Cardinal, 2001; Ohly et al., 2006; Gilson et al., in press; see also the review in Adler, 2007). As regards formalization more specifically, a meta-review did not find it to have a negative impact on innovation (Damanpour, 1991), and some recent studies have even found a positive relationship (Andrews and Smith, 1996; Johnson et al., 1997; Mellor and Mathieu, 1999; see also Hargadon and Sutton’s (1997) claim that following routines such as ‘brainstorming routines’ played a crucial role in IDEO’s success with radical innovation).

The role of procedures and stable behaviour patterns in innovation

225

Such contradictory evidence raises the question ‘what role, if any, do standard operating procedures and stable behaviour patterns play in repeatedly generating innovations?’ Because this is an empirical question of exploratory nature, we tackle the question with a case study in the automotive industry. Based on the empirical findings, we advance the proposition that under certain circumstances, standard operating procedures and stable behaviour patterns have the potential to contribute to and foster even radical innovation. The chapter aims to establish this empirically grounded proposition and feed it into the literature concerned with innovation management, thereby contributing to casting more light on the untapped potential of standard operating procedures and stable behaviour patterns in fostering exploration processes and even radical innovation. Such potential can have significant implications both for practice and for frameworks of innovation management and new product development. Importantly, the idea of routines as being negatively correlated with innovation can lead to neglecting powerful innovation tools and dismissing non-trivial potentials and levers for innovation, including radical innovation.

2.

ORGANIZATIONAL ROUTINES

As set out above, our interest in this chapter is in understanding the potential role of one endogenous factor of innovation, the routinized accomplishment of innovation tasks (brought about by such means as standard operating procedures). The concept of organizational routines is helpful for capturing how an organization typically accomplishes its tasks (Nelson and Winter, 1982). Where a typical response to a problem exists, innovation tasks are solved by stable behaviour patterns. The term ‘stable behaviour pattern’ refers to accomplishing a particular task in always the same way.2 The idea of stable behaviour patterns for accomplishing innovation tasks therefore points to the hypothesis that the sources of repeated innovation (that is, stability in outcomes of innovation processes, possibly even radical innovation) can be found in stability in the innovation process (how innovation tasks are accomplished). Standard operating procedures are possible sources of the stability of such behaviour. They are not the only sources of stable patterns of behaviour, however. Scholars have also identified cognitive factors, individual dispositions, and other factors as sources of the stability of behaviour (Cohen et al., 1996; Feldman and Pentland, 2003; Hodgson, Chapter 3, this volume; Knudsen, 2008; Nelson and Winter, 1982; Schulz, 2008). We emphasize standard operating procedures here because such an emphasis is highly consistent with the emphasis

226

Organizational routines and organizational change and innovation

on procedures in the new product development literature (described in more detail in section 4). At this point, a few words on the concept of ‘organizational routines’ are in order. Because authors have used the term in diﬀerent ways, it is important to be clear about our use of the term here. A widely diﬀused interpretation of the term ‘organizational routines’ is to consider stable behaviour patterns ‘organizational routines’. Some authors, however, argue that organizational routines should refer to the generative level, that is, the level that causes (stable) behaviour, rather than the level of behaviour itself (for example, Pentland and Rueter, 1994; Knudsen, 2008; Hodgson, Chapter 3, this volume). We side with the latter view. At the generative level, two candidates of sources of behavioural stability have been identified: standard operating procedures and individual dispositions. For reasons explained in section 4, our attention in this chapter will be on standard operating procedures. For the sake of clarity, in what follows, we identify which of the three notions (stable behaviour patterns, standard operating procedures, or dispositions) we refer to. In this way, we discuss what many authors consider ‘organizational routines’, but use a more precise terminology that allows better framing of our contribution.

3.

THE LITERATURE ON ORGANIZATIONAL ROUTINES IN INNOVATION

How to explain why some firms manage to maintain a repeat series of radical innovation, where others settle into merely incremental innovation, or no innovation at all? Taking the vantage point described above invites considering standard operating procedures and stable behaviour patterns as important parts of the answer. As mentioned, not only are they often not seen as part of the answer, but rather, as part of the problem. Because of the supposed contradiction between routines and radical innovation, the question of the role of standard operating procedures and stable behaviour patterns has not yet been answered fully. A few scholars have, however, argued that considering routines as blocking radical innovation is not a completely correct picture of the role of routines in innovation. Rather, these scholars see routines fulfilling a role in innovation activities that has a positive impact on the outcome of such innovation activities. Nelson and Winter (1982: 129), for instance, discuss the ‘way in which the routine functioning of an organization can contribute to the emergence of innovation’ and refer to innovation routines as ‘routines for the support and direction of [firms’] innovative eﬀorts’ (1982: 134). Amongst other things, Nelson and Winter arrive at the conclusion that ‘there is, however, more to be said

The role of procedures and stable behaviour patterns in innovation

227

about the relations of routine behavior and innovation than to observe that these concepts are commonly (and appropriately) regarded as opposed ideas’ (Nelson and Winter, 1982: 129), and that at the very least, ‘important qualification[s] to the general notion of an opposition between routinization and innovation’ need to be made (Nelson and Winter, 1982: 131). Almost none of such qualifications have come forth, however. Specifying the relationship of organizational routines and innovation has remained an almost completely neglected task.3 This fact led the late Keith Pavitt, one of the leading scholars of the economics of innovation, to dedicate his keynote lecture at the 2000 International Schumpeter Society meeting to ‘innovating routines’. On this occasion, he emphasized that ‘we still have only a very hazy idea of what innovating routines are in practice’ (Pavitt, 2002: 117). Pavitt (2002: 118) also argued that knowledge of innovating routines – especially in large firms – deserves greater attention because it helps identify ingredients for the successful management of innovation; because it allows a more realistic interpretation of what managers actually do in a messy and changing world; and because there are opportunities of successfully combining the new theoretical concepts of innovating routines with rich bodies of empirical evidence on what happens inside the innovating firm. Giving ‘more operational content to the notion of innovating routines’ (Pavitt, 2002: 129) was called for. The call has not been heeded very much. Recently, a few attempts have come under way to explore the role of routines in innovation (Turner, 2003; Turner et al., 2004). The idea of investigating the role of routinized accomplishment of innovation tasks thus has been around for some time. Only recently has it picked up some speed.

4.

THE NOTION OF ORGANIZATIONAL ROUTINES IN NEW PRODUCT DEVELOPMENT

Dealing with empirical description and analysis of new product development processes, it is obvious to turn to the new product development literature.4 According to the new product development literature, one of the main drivers of successful new product development (in terms of lead time and cost) is proper project management of the new product development process (Cooper and Kleinschmidt, 1995; Kahn et al., 2006). The reason is that for complex products in particular, coordination is crucial for achieving brief lead times and low cost. These are achieved through concurrent engineering and good project management. Successful project management is assured mainly by two means: organization structure (crossfunctional teams, heavyweight project teams, matrix structures and so on;

228

Organizational routines and organizational change and innovation

Wheelwright and Clark, 1992), and project management procedures and standards (such as stage-gate models and concurrent engineering practices; Cooper, 1990; Griﬃn and Hauser, 1996). The new product development literature thus emphasizes a formalized new product process, procedures, and carrying out innovation tasks in stable, routinized ways (Kahn et al., 2006). As much empirical research of innovating firms shows, this is actually what many firms do, structuring and systematizing the new product development process with the help of standard operating procedures and stable behaviour patterns.5 The new product development literature has, thus, greatly emphasized standard operating procedures in the innovation process. Such an emphasis is, obviously, consistent with the emphasis on standard operating procedures in March and Simon (1958 [1993]) and Cyert and March (1963 [1992]), and with the ideas on innovation routines proposed by Nelson and Winter (1982) and Pavitt (2002).

5.

ANALYTICAL PERSPECTIVE AND METHOD

As Pavitt (2002) argued, the analysis of experimentation and its role in innovation is an appropriate starting point for casting light on innovation routines. As Pavitt (2002) also mentioned, one of the challenges for innovating firms is to adapt to the possibilities oﬀered by reduced costs of technical experiments provided by the possibility of applying virtual simulation technology for experimentation. For our empirical work, we have therefore selected experimentation. Nowadays, experimentation is increasingly carried out with virtual development tools. As we shall see, virtual simulation in experimentation is routinized, and is subject to procedures. Experimentation is a key task of the product development process. It is a source of learning about system characteristics that are diﬃcult to calculate in a deductive manner (for instance, noise, material fatigue, tightness of seals and so on), and about quality problems, fulfilling legal requirements, and the robustness of tooling or assembly. A new car model, for instance, is developed over a period of roughly (up to) four years. During that time, hundreds of subsystems are developed, often in cooperation with suppliers that need to be integrated into a vehicle that attains certain performance objectives. Engineering solutions, developed by applying codified engineering knowledge (for example, for designing a suspension), need to be tested before they enter production. Engineering solutions, hence, are developed by iterated trial and error, with experimentation being particularly important because it serves to guide the development eﬀort. Traditionally, experimentation was performed on physical prototypes. Prototypes of entire vehicles in particular were extremely costly (about 50 per cent of the overall

The role of procedures and stable behaviour patterns in innovation

229

development cost for the type of projects under investigation). Producing them (in crafts manner) is a highly time-consuming process. Nowadays, virtual development tools (including virtual representation tools and virtual simulation tools) allow for experimentation in a virtual environment, that is, simulating the interaction of components that are described in CAD files. We thus focus on the use of virtual simulation in experimentation. In order to explore the research question, we carried out observations both on the process of setting up the virtual development tools and on their implementation in practice. The observations referring to the use of development tools, on which this chapter focuses, cover the time lag from vehicle concept definition to style freezing and production tooling, encompassing new product development (from now on ‘NPD’) phases in which the ‘exploration’ attitude is prevalent and phases that are characterized by the prevalence of an ‘exploitation’ approach. This approach, the nature of the research question and setting, made an exploratory case study method the most appropriate method (Eisenhardt, 1989; Pettigrew, 1990; Yin, 1994). The empirical material we present is part of a broader project on the NPD process in the auto industry started in 1997. In this chapter, we present a case study carried out between October 2002 and October 2004. We studied the aforementioned experimentation process at a research and design centre in which both fundamental research (such as research on new materials) and applied development activities (such as the development of a suspension for a specific vehicle) are carried out. The research centre chosen for the investigation has recently begun to play a central role in the process of ‘virtualization’ of the auto maker’s (from now on ‘OEM’ for ‘original equipment manufacturer’) new product development process. It counts around 750 employees and was established as a ‘green field’ research centre close to one of the OEM’s plants in 1988. The low average age of its engineers and their high standard university training made the research centre an ideal place for the OEM to invest in new technologies and methodologies for product development. Field observations have been carried out both on the process of setting up the virtual development tools and on their implementation in practice. It is important to underline that these two phases are clearly separated and diverse. In particular, the observations referring to the use of development tools covers the time lag from the vehicle concept definition to style freezing. We believe this setting is ideal given that in this phase, virtual tools have to prove they can comply with strict time and cost targets. As a consequence, our insights apply to the stage of product developments closer to the product launch on the market (from approximately five years to the launch). Three data collection methods were used. (1) Archival sources. About 2000 pages of oﬃcial company documents (norms and procedures) were

230

Organizational routines and organizational change and innovation

analysed. (2) Extensive semi-structured interviews. More than 60 hours of interviews were carried out. Interviews involved the head of the vehicle department; the head of project management for vehicle department; the vehicle development function manager; the head of human resources; the chief engineer of ‘packaging’ (the task we focused on); one ‘packaging’ team leader; three ‘packaging’ team members; the project manager. The project was the most important one the research centre was carrying out at the time of observation. Finally, (3) one researcher spent 10 months working at the research centre. During this time, he regularly took part in meetings, participated in some of the tasks, provided support, and so on. The study is subject to limitations. This chapter presents a preliminary attempt to pursue the idea of innovation routines by an exploratory case study. While the setting and innovation task we have chosen is a good example for innovation tasks (also, the firm we studied is clearly within that realm), it also has its specificities. Most importantly, in our case, the context was highly technical (automotive industry, mechanical design, virtual simulation). In this context, procedures turned out to be highly influential. One of the reasons might in fact be that the context was highly technical, and only by standardizing the use of the tool within the firm can the tool be eﬀective at all. For other tasks, or in other contexts, the sources of stability might well be diﬀerent ones, for instance, cognitive factors, interlocking behaviours, and so on.6

6.

EMPIRICAL FINDINGS

We now turn to describing the empirical findings, structured in three steps: how virtual development tools are set up, how they are used, and how their use aﬀects innovation outcomes. In order to identify the latter, we identify opportunities provided by virtual simulation tools, and organizational mechanisms used for leveraging such opportunities. 6.1

How Virtual Development Tools are Set Up

In our observations of experimentation by virtual simulation tools, one of the findings that struck us was how much attention and eﬀort the firm spent on developing procedures for using the simulation software.7 A whole department was, in fact, dedicated to writing the procedures on how to use the virtual simulation tools. We found a plethora of procedures and norms pertaining to the experimentation task. In setting up virtual development tools, most energy is focused on developing procedures for applying the virtual development tools. Having scouted simulation

The role of procedures and stable behaviour patterns in innovation

231

technologies available on the market (software development is always outsourced), research centre staﬀ start developing procedures for the use of virtual development tools (from now on ‘procedure’) by mathematically modelling the behaviour of a vehicle that is already on the market, and on which it is possible to test the real performances. The input data on which the procedure is built, hence, are real design data coming from an existing vehicle model. The first release of the procedure is tested by comparing the results of the vehicle performance performed on the real vehicle in a full-scale simulation, with the output of the virtual simulation based on the same input data. The procedure is refined on the basis of the comparison between the results obtained from the virtual simulation and the results obtained by the physical experimentation on the real vehicle. It is worth noting that the reliability of the virtual tool has to approximate 100 per cent in order to get a first validation. At this stage there is a first release of the procedure, which allows reliable simulations on the vehicle that has served as the test object (data coming from virtual and real experimentation converge). The problem at this stage is that the procedure is reliable for a posteriori validation of the vehicle that is already on the market, but is not necessarily able to predict the performance of a completely new vehicle. In other words, the procedure has to prove to be generally applicable. The final validation of the procedure, hence, is released only when the new procedure is tested on another real case. Usually, the new comparison is performed with a second vehicle already on the market, which is considerably diﬀerent from the one on which the procedure has been previously defined. The procedure is validated when it allows simulation results to be generated that are reliable and general enough to be applied in simulating vehicle performance of a future model: an acceptable threshold for this purpose is a 1/– 3 per cent gap between simulation output and real performance. The procedure development just described may last from 12 to 24 months, varying according to the complexity of the performance to be simulated. This time span attests to the importance of the procedures for the successful use of virtual development tools, in particular to allow the designers to predict the behaviour of the vehicle before physical tests are performed. 6.2

How Virtual Simulation Tools are Used

Procedures for virtual development show the user (the OEM development team) how to use virtual simulation tools, extracting information previously obtained only after physical experimentation. In conjunction with the simulation tool, the procedure provides: (i) a description of how to design the object on whose performances the simulation will be modelled;

232

Organizational routines and organizational change and innovation

(ii) calculus norms; and (iii) the error rate that is considered acceptable in the simulation (that is, the gap between the output of the simulation and what will happen in reality). These three elements allow the development team to benefit from software simulation tools, and to eﬀectively apply virtual technologies/tools in product/process development. This is because procedures (1) integrate design norms and procedures and (2) enable their application to virtual representations. ‘Design norms’ specify how a certain object has to be designed and engineered, and ‘procedures’ specify the flux of activities to be performed and the task partitioning (‘who does what’). In considering the role of procedures, it is important to remember that only the correct use (correct input procedure) and interpretation (correct analysis of the simulation results to predict the real behaviour of the vehicle) of the procedure will allow a successful vehicle design that will achieve a certain reliability in the virtual experimentation of the vehicle performance. In other words, the procedure will enable the engineers to use the simulation tool in a way that is reliable enough in simulating vehicle performances before physical experiments are carried out. The procedure shows the user (the OEM development team) how to use virtual simulation tools, anticipating information previously obtained only after physical experimentation. Procedures are used in the early stages of product development both by members of vehicle development teams and members of functional units of the OEM who then provide simulation results to vehicle development teams (these phases can be framed approximately between 60 and 40 months before vehicle launch on the market). The full range of procedures for virtual development should realize the coordination necessary to codify information on the various components and systems of the vehicle into a digital format that, through the company PDM (Product Data Management System), should be the input data for all the simulation activities during the early phases of vehicle design (for example, the computer-aided engineering (CAE) and computer-aided manufacturing (CAM) teams should use the same design input, stored in the PDM, to perform their simulations). This shows how, in the case of virtual simulation, procedures are increasingly being incorporated in the software. 6.3

How the Use of Virtual Simulation Aﬀects Innovation Outcomes

We address the research question by following the structure described in Figure 10.1, presenting the argument along the steps described there. The framework focuses on the way in which virtual development technologies were used. In doing so, our interest is in capturing what is systematic about their use (rather than one-oﬀ events). This is why we emphasize the concept of organizational routines in order to describe how the firm

The role of procedures and stable behaviour patterns in innovation Opportunities provided by virtual simulation tools

Figure 10.1

Organizational mechanisms (incl. procedures) to leverage the opportunities

233

Impact on innovation outcomes

Framework on eﬀects of applying virtual simulation technology

typically and systematically uses virtual development tools. As described in section 2, behind the broad concept of ‘organizational routines’ is the following idea: performance is generated by behaviour. When such behaviour is systematic (that is, tasks are accomplished in the same way every time the task has to be accomplished), such behaviour takes the form of recurrent behaviour patterns. These patterns are influenced and their stability is generated by procedures.8 Procedures channel and guide the emergence of particular recurrent behaviour patterns. This is why we apply a perspective that sees the performance of virtual development tools as described in Figure 10.1: technology in combination with organizational mechanisms, including procedures, generate performance outcomes. Such a perspective is consistent with the idea that where innovation tasks are always accomplished in the same way, the performance outcomes will be generated by stable behaviour patterns. 6.3.1 Opportunities provided by virtual simulation tools In this section, we prepare the ground for our argument by describing what opportunities virtual development technology provides.9 Such technology provides the following opportunities that were not available before. It helps engineers observe phenomena which are much less readily observable otherwise. The most straightforward example is a crash test. One of the big limits to observability in a physical crash test is that it is very diﬃcult to observe the impact of the crash on many of the internal parts of the car. The impact needs to be reconstructed ex post by analysing deformations of the crashed prototype, clearly with a huge loss of valuable information. Alternatively, sensors and high-speed cameras can be fitted internally, with the problem that they, too, will be destroyed. In short, the direct limits to observability are substantial in physical crash tests. There are indirect limits, too, which we explain in the following. Virtual simulation tools enable almost infinite iterations of the same experiment, due to the higher speed and lower cost of running each iteration. This is remarkable in particular when taking into account the fact that for some kinds of experiment, the number of runs can be very low. As it happens, this is the case in the example we have chosen. Doing a physical frontal crash test means crashing a physical prototype. Developing a physical prototype takes about two months, with costs between $250 000

234

Organizational routines and organizational change and innovation

and $400 000, including physical tests, planning and execution; on the other hand a digital model for crash analysis takes less than one man-day, and the whole crash simulation takes no more than two days, even outside working hours, as the presence of the operator is not necessary. With virtual simulation, one can go from a handful of runs to many thousands of runs. Importantly, virtual development tools also allow isolating one parameter in each run (they create an almost ‘laboratory-type controlled environment’). Given the fact that there are at least 400 components (our interview with a product development engineer) that aﬀect such a crash test, it is impossible to isolate precise cause–eﬀect relationships among all the parts without repeating the experiment several times. That possibility is only given with virtual simulation tools. Thanks to the high number of runs and the possibility of isolating one parameter in each run, phenomena become observable that were simply hidden from view before. Because they are digital tools, virtual development tools automatically leave traces of all runs and their results. The traces left in the system provide a strong organizational memory of all the experimentation runs. As Schön (1983: 157–8) explains, the residual traces left by using virtual simulation tools are stable, and thus, the designer can examine them at leisure while the pace of action can be varied at will. Due to the high speed and low cost, many hypotheses on the causal relationship between the design characteristics can be tested. High numbers of runs means huge possibilities of testing many hypotheses, including unusual hypotheses that are not constrained by the logical bounds of the premises one starts from. It thereby enables non-conservative design, which is important in order to achieve distinctive new designs (Wheelwright and Clark, 1992). In testing a large number of hypotheses, because of the speed and cheapness of a run, it is possible to test ‘random’ hypotheses in the sense that one changes, for instances, parameters where one does not have any particular reason for believing they should have an impact. (Random screening of chemical compounds in pharmaceuticals is a case in point.) That also means one can shed constraints from the logical bounds of the premises one starts from, or from experience and history (Becker and Zirpoli, 2006). 6.3.2 Organizational mechanisms for leveraging opportunities As already mentioned, the opportunities oﬀered by virtual simulation technology are levered by organizational mechanisms. Organizational mechanisms can guide people to realize these opportunities systematically, that is, through stable behaviour patterns. As described above, procedures for using virtual simulation tools are one particularly prominent element of guidance and leverage in our case. (Procedures also have an important

The role of procedures and stable behaviour patterns in innovation

235

role in giving rise to stable behaviour patterns.) As mentioned above, much eﬀort was spent on writing such procedures. Because of the prominence of procedures in the empirical case, we focus on describing the impact of the procedures on accomplishing the experimentation task. In our empirical research, we have found that procedures assure a number of preconditions for successfully using the physical technology in accomplishing the experimentation task. These can be clustered in three groups: Procedures provide control of the innovation process Successfully applying virtual development tools requires the collection and inputting of (all) the required data, as well as the capability to interpret the output of the virtual development tools and link them to other development tools. We focus on entering data. In order to benefit fully from the use of virtual simulation tools, product developers are often forced to input and process data in a standardized way. Such standardization leads to better control of the design process both ex ante, via the setting of standard procedures for designing new components or systems, and ex post by the monitoring of the data. Only because information is entered in a standardized way is the output of the system reliable. The reason is that only the standardized input of data allows for the standardized application of tools. Procedures accomplish integration of the innovation process Procedures for virtual development accomplish integration between diﬀerent elements (more below) in the experimentation process. Integration (into a smoothly coordinated process) being one of the key challenges in developing complex products, this is an important eﬀect of procedures for virtual development. Such procedures accomplish integration in at least three diﬀerent instances: First, they integrate design norms (‘how’ a certain object has to be designed and engineered) and procedures (the flux of activities to be performed and ‘who does what and when’), and thus enable their application to virtual representations. Second, they organize the integration of physical and virtual simulation results. As mentioned briefly, for calibrating and validating the simulation model, the results of the virtual simulations have to be compared with physical experimentation results. In order to have both physical and virtual simulation results available, one calibrates the model by simulating a car that is already on the market, and for which real data and physical experimentation results are available. A complex iterative comparison between physical and virtual experimentation is then required in order to embed the technical know-how of the company in the virtual tool. The point is that for virtual simulations to be reliable, virtual predictions need to be tested against and calibrated with physical

236

Organizational routines and organizational change and innovation

experimentation results. It is a necessary requirement for arriving at reliable virtual simulation results. Nothing in the tool itself, however, assures that such a check will take place. How this check is organized is, to a large extent, described in the procedures. How well the use of the virtual development tool leads to a reliable outcome (which requires iterations between physical experimentation results and virtual experimentation results) will be decided by the procedures, and how they are implemented. Where they are successful in organizing such an iterative comparison, they accomplish the integration of physical and virtual experimentation – not an obvious feat. Finally, due to the standardization that the procedures for virtual development introduce, they also integrate diﬀerent development tools, for instance the product data management system (PDM), the CAD/CAM systems, and so on. Procedures accomplish such integration by specifying which data is to be used where, how it is to be transformed, saved, and so on. What is important here is that all these specifications are standardized, that is, contribute to inter-operator reliability. The integrating eﬀect, in all three cases, is driven predominantly by the procedures. The reason is that the procedures that concern integration are often embodied in the software (such procedures do, for instance, regulate how to integrate various sources of data, communication between various databases, and so on). Procedures guide attention during the innovation process Procedures also have eﬀects on the people that carry them out. One such eﬀect is to guide people’s attention (March, 1994; Ocasio, 1997). By specifying which kind of data to input, procedures focus actors’ attention on certain types of data. Similarly, and to an even higher degree, in specifying how to interpret results, the procedures guide people’s attention in a particular direction. For instance, they provide hints at what aspects of the results are especially important, what are the critical thresholds, and so on. These examples also demonstrate to what extent the benefits of virtual simulation tools really depend on the procedures for how to use such tools. The procedures do indeed act as an important lever. In this case, too, the eﬀect is due to the procedures themselves. It is the procedures that focus attention, for example, by specifying which variables need to be included in a particular calculus. A necessary condition, however, is that the procedures need to be perceived by employees (otherwise, there is no attention to guide). So far, we have focused our discussion on procedures. In accordance with our framework, as described in Figure 10.1, we have considered how the combination of technological tools and procedures for how to apply them can generate or foster stable behaviour patterns in which the technological tools are applied for solving problems. The causes of stability of such behaviour patterns itself is not the main focus of the present chapter,

The role of procedures and stable behaviour patterns in innovation

237

however, and we leave this issue for further investigation at this point. In the section that follows, we rather shift our attention to considering the consequences of solving problems in stable behaviour patterns. According to the wide-spread idea, the expectation is that stability in accomplishing tasks will lead to enhanced eﬃciency in accomplishing the task, but not to radical innovation – and perhaps even to systematically excluding radical innovation. As we will argue in the remainder of the chapter, there is more to the matter than that. 6.3.3 Impact on innovation outcomes In our analysis of the impact of the procedures, it emerged clearly that the procedures introduced possibilities that the physical tools themselves did not already provide. However, procedures (and other organizational mechanisms such as governance mechanisms and incentive structures) will only provide such eﬀects to the extent they are actually followed, and become recurrent behaviour patterns. Where people ignore the procedures, or interpret them in changing and erratic ways, no stability will be generated. (The possibilities for not following the procedure can be limited, however. In the example of data entry, for instance, one can provide only one format to input data.) In this chapter we focus our attention on the consequences of the emergence of stable recurrent behaviour patterns on innovation outcomes. Our field work provides empirically grounded reason for considering the role of stability as a prerequisite for successful innovation more carefully. More specifically, in the case we studied, we found stability to lead to eﬃciency and control of virtual simulation. Importantly, however, such stability in the way that tools are applied does not merely lead to increased speed and lower cost of virtual simulation. It also leads to regularity in data input and compatibility (between diﬀerent operators). These eﬀects are to be expected, and are indeed often attributed to virtual simulation tools (see Thomke, 1998a, 1998b; Thomke and Fujimoto, 2000). Sometimes, they are the only eﬀect of virtual simulation procedures that is perceived. There are a number of other important eﬀects of stability, however. It seems that by overlooking those, one would miss important eﬀects of procedures (which contribute to accomplishing innovation tasks such as experimentation in a stable way) on the outcomes of innovation tasks. For one, regularity in data input and compatibility allows better monitoring (as standardization will take out much ‘noise’ and make it easier to identify meaningful patterns in the data). Second, stability (for example in how data are entered, interpreted, and so on) also establishes a base-line against which one can compare. This is a precondition for learning. Third, stability also enhances the non-trivial task of integration, as pointed out

238

Organizational routines and organizational change and innovation

above. The role of procedures in coordinating the systematic confrontation and iteration between virtual and physical experimentation achieves integration between virtual and physical experimentation. This point illustrates how procedures play an important role for the quality of the results attained by using such simulation tools. Fourth, stability provided by the procedures helps guide attention. This feature can increase consistency in interpretation and common frameworks. To appreciate this point, remember that the output of the simulation tools is just numbers, that is, raw data. Before such data can be useful in drawing conclusions for solving design problems, it has to be transformed in information and knowledge (the diﬀerence between them is contextualization, cf. Bohn, 1994). This can be a lengthy process, in the course of which individual-level factors such as individual cognition lead to diﬀerent results – people might arrive at very diﬀerent conclusions regarding what the same data mean. In guiding people’s attention, procedures create some degree of consistency (common framework) that will foster an interpretation that is easy to communicate and act upon. Procedures guide attention to a selection of variables, and reasons why they are to be considered particularly important. They can contribute to consistency in people’s mental models (even though they might not bring about consistency by themselves). Such consistency can, however, be reinforced and fostered by the mutual feedback people get in actual stable behaviour patterns. All those eﬀects of stability, provided by procedures, are substantial and important, and ignoring them would mean to neglect some aspects of how innovation tasks are accomplished that seem to have important eﬀects on innovation outcomes. The most important and counter-intuitive point, however, is that stability provided by the procedures can also enhance exploration. To see why, it helps to consider innovation as a search process (Winter, 1971; Winter, 1984; Katila and Ahuja, 2002; Fleming and Sorensen, 2004), for instance for solutions for technical problems, or for new combinations (Schumpeter, 1934). Search processes have been explored by NK models (Levinthal, 1997), where actors search in a space of N elements with K interdependencies. Particularly in large spaces, how the search process is organized and guided has a huge impact on what part of the search space will be searched and thus, which solutions will be discovered. Research on search processes in the NK modelling tradition (Levinthal, 1997) has also shown that the search path is important because it is possible – and indeed, under many circumstances likely – that actors will settle on sub-optimal points in the landscape (for instance, a new product that does not represent the best possible combination of attributes as valued by the customer). Influences on the direction of search, the granularity of search (at what level of detail one searches), and the generation and processing of feedback are all examples

The role of procedures and stable behaviour patterns in innovation

239

of factors that have an important influence on search processes and their outcomes. Procedures seem to be particularly important levers by which search processes are influenced, guiding the direction, granularity, and the generation and processing of feedback. This is in fact precisely what we have seen in the case study. Exploring the role of procedures on search processes is also important because (a) they are so prominent in a new product development context, and (b) they are one of the most prominent variables for describing organizations (formalization). The influences of procedures on search processes (by stabilizing the particular way in which the experimentation task is accomplished) can explain why following procedures can lead to radical innovations. Above, we mentioned that virtual simulation enables the number of experiments to be increased and their cost to be lowered dramatically, and also enables the testing of ‘random’ hypotheses. These features provide the possibility of an increased number of surprising results.10 Surprising results represent opportunities for reflecting, that is, for attempting to make sense of the surprise, for re-setting decisions about the search process, for adapting the search process and so on. Procedures for virtual development have perhaps the most immediate influence on whether such an opportunity will be systematically used or not. For instance, the procedures can require that a process of reflection be triggered and documented. A surprising result can also be used to justify going forward with an unconventional idea, such as a product design (as we actually observed in the case). To the extent that the occasion to reflect is used, reflection can lead to redefining the search path, for instance, by focusing on a diﬀerent part of the search space, resetting the direction in which one searches, shifting to a diﬀerent level of granularity, and so on. In conclusion, the list of possible eﬀects of procedures is long. Stability provided by the procedures (and their recurrent implementation) is not just responsible for an eﬃcient application of the virtual simulation tools and for increasing eﬃciency (lead time, cost). Rather, it can have other important eﬀects on the outcomes of innovation tasks that go beyond eﬃciency. Taking those into account can substantially strengthen our understanding of how to organize innovation, including why some firms manage to persist in generating radical innovations.

7.

STABILITY AND CHANGE: THE ROLE OF PROCEDURES AND STABLE BEHAVIOUR PATTERNS IN ENHANCING INNOVATION

As underlined above, the procedures used in order to enhance the implementation and use of virtual development contribute to trigger and guide

240

Organizational routines and organizational change and innovation

recurrent behaviour patterns. In this concluding section, we will address the main question that we have raised in the course of the chapter: what is the role, if any, of standard operating procedures and stable behaviour patterns in repeatedly generating innovations? Our main findings are two: (1) Stability (induced by procedures) is a prerequisite for being able to innovate repeatedly. (2) Given such stability, procedures can also trigger exploration, not just exploitation. Procedures for virtual experimentation provide important prerequisites for successful product development. For example, in the case of virtual tools, the possibility to trace design releases and their impact on vehicle performances is a prerequisite for being able to reflect in detail, and reconstruct what has happened in earlier experiments, that is, for being able to trigger a process of reflection and innovative problem solving. Exactly the same mechanism is at work when the procedure for inputting and interpreting data from a simulation run is defined. Standardized data input is a precondition for consistent and comparable experiments. A somewhat consistent interpretation across the firm is a precondition for the possibility that conclusions from the experiments can accumulate and build on previous ones. Procedures also provide the stability that enables varying just one variable at each run and thus have a base-line for comparison, a crucial prerequisite for learning. Similarly, in companies such as IDEO, the stability of the ways in which people interact with potential users allows them to be systematically innovative because a common language and interpretation framework is shared by members of the firm. The insight here is that in diﬀerent ways, stability (such as induced by procedures) can be a prerequisite for being able to innovate. Being able to pick up on puzzling results, for instance, is only made possible by the stability introduced by procedures regarding data input. The highly standardized way (driven by the procedures) in which engineers carry out the experimentation task ensures a base-line against which results are comparable, and surprising results stick out. The second non-intuitive finding is that procedures can also be the trigger that realizes possibilities for exploration oﬀered by virtual experimentation tools. Once there are comparable simulation results, it is always possible for an engineer to start wondering about a puzzling result. Procedures can make it more likely that a puzzling result will actually give rise to reflection. Procedures can trigger exploration, for instance, by guiding people’s attention to certain parameters, or by requiring a discussion, meeting, or report on outliers. It is important to note that whether actors actually pick up the possibility of, for instance, reflecting on puzzling results and engaging in exploration, is always at the actors’ discretion. Given a certain predisposition of actors to do so, procedures can, however, trigger this predisposition to be realized.11

The role of procedures and stable behaviour patterns in innovation

241

Note that the first feature, stability, also has the consequence of enhancing the eﬃciency of the testing task. The second feature, on the other hand, explains how the same procedures can, under some circumstances, lead to exploration. It now becomes clear why and under which circumstances development routines can be reliable and innovative at the same time. This will be the case if the behaviour patterns that are stabilized also include the possibility of changing search paths or redefining the search space. Our findings therefore point to procedures as one element for closing the gap identified by Adler (2007), that is the lack of a theoretical characterization of a form of organization capable of ‘squaring the circle’ of innovation and eﬃciency. Our findings gain significance in relation to extant research on the sources of exploration. As Greve (2007: 25) argued, we need to expand the search of variables that shift between exploration and exploitation, and extend it to also identify factors that lead to more or less explorative innovations. In this chapter, we focus on scrutinizing the organizational conditions within which agents (which have an individual disposition to explore) are set. We take agents’ dispositions to explore as given and focus on what diﬀerence diﬀerent organizational settings make for the extent to which agents in that organizational setting will engage in exploration. The organization of the search process thus can systematically favour the ‘explorative’ mind (see also Levinthal and Rerup’s (2006) systematic ‘processes for sustaining mindfulness’). Such a finding is plausible and consistent with the result of Damanpour’s (1991) meta-analysis that many drivers generating radical and incremental innovations are the same.

8.

CONCLUSION

In looking forward, we would like to highlight two points. First, the key point of our empirically grounded argument is that at least under certain circumstances, the routinized accomplishment of innovation tasks can be an endogenous source of innovations, including radical innovations. Routinized accomplishment of innovation tasks can contribute to fostering exploration because it has an impact on the search process. In this chapter, we have sketched a possible starting point for future research that probes this idea. Much remains to be done in order to further corroborate this argument and fill in the details. At this point, we just want to point to an interesting parallel. The proposition just made is similar to the proposition, recently advanced, that routines can be an endogenous source of organizational change (Feldman and Pentland, 2003). We consider this proposition in the specific setting of innovation, and as having an eﬀect on the search process, rather than on organizational change.

242

Organizational routines and organizational change and innovation

As to the second point to highlight, the case indicates that the eﬀects of routinizing experimentation are not straightforward: while procedures easily seem to enhance exploitation (more eﬃcient project management), they can also enhance exploration. Which of the two they will enhance is influenced by, for instance, the content of the procedures. Procedures can push towards a standard (conventional) use of virtual tools (substituting physical experimentation, in order to be more eﬃcient). They can also, however, induce a use of innovation tools that is more conducive to exploration (taking advantage of an improved technical memory and the possibility of almost infinite experimentation runs to explore unconventional solutions). In consequence, procedures (and the recurrent behaviour patterns they engender) seem to hold one of the keys both for providing the prerequisites for exploration that leads to radical innovation, and for ‘switching’ from exploitation and incremental innovation outcomes to exploration, with potentially radical innovation outcomes. Procedures therefore can be levers for managers in designing organizational mechanisms that push in one direction or another. In firms where the innovativeness and originality of design is a strategic priority, the design of the procedures can push engineers to use virtual tools to test non-conventional solutions. Their design activity will still be constrained by the procedures for virtual development, but their attention will not be devoted exclusively to standardization of design spaces and domains. It is interesting to note that the control aspect of virtual simulation tools will accrue almost by itself, in the light of the fact that procedures for virtual development are required in order for virtual simulation to provide reliable results (this applies, of course, also to the procedures in companies such as IDEO and Apple). Whether the use of virtual simulation will lead to innovative designs, on the other hand, depends almost entirely on the alignment and fine-tuning of the procedures to the tools and the organization structure and management system. A final remark concerns a question raised by the insight that stability induced by procedures (in our example) is a source of eﬃciency enhancements in the new product development process but also (if to a lesser degree and only by providing the preconditions) of the possibility to engage in exploration. In the light of this insight, the supposed contradiction between either eﬃcient innovation processes or radical innovation does, at least to some degree, appear fictitious when we start considering the micro-details. So does the conviction that procedures and the recurrent behaviour patterns they engender are not conducive to innovation.

The role of procedures and stable behaviour patterns in innovation

243

ACKNOWLEDGEMENTS Authors’ names are in alphabetical order because they have contributed to the chapter equally. For purposes of formal assignment, Markus Becker wrote sections 2 and 3. Francesco Zirpoli wrote sections 4, 5 and 6. Section 1, 6.1, 6.2, 6.3, 7 and 8 were written jointly. We are grateful to Paul S. Adler, Francois de Vaujany and Carlo Salvato for comments on earlier versions of this paper. All errors and omissions are our own responsibility. Support from the Agence Nationale de Recherche, France (‘Jeunes Chercheuses et Jeunes Chercheurs’ programme, grant no. JC05_44029). and the Italian Ministry of Higher Education and Research (MIUR) under the PRIN 2004 programme (project 2004135057_004) is gratefully acknowledged.

NOTES 1. 2.

3. 4. 5.

6.

7. 8. 9. 10. 11.

With the help of standard operating procedures and stable behaviour patterns, for instance. There is a consensus in the literature that the concept of organizational routine refers to the collective level, while the concept of habit refers to the individual level (Dosi et al., 2000). The term ‘recurrent interaction pattern’ denotes such collective, multi-actor patterns of behaviour precisely. We here refer to the term ‘organizational routines’ more specifically. For one of the few articles explicitly discussing innovation and routines see Benghozi (1990). The paradigmatic journal of this literature is the Journal of Product Innovation Management. Note that in the following, we use the term ‘product innovation’ to refer to the literature stream. We do not mean to exclude process innovation. A glance through the latest volume of the Journal of Product Innovation Management or R&D Management will lead to empirical research that describes how firms implement stage-gate processes, ‘funnels’ in the development process, and accomplish the innovation task in other routinized ways. Another limitation is that we have only explored the role of procedures. However, a setting with the specificities just described makes it all the more surprising that we find, as argued below, that routinizing experimentation can also lead to exploration – after all, both procedures and information technology are usually associated with increased eﬃciency, and thus exploitation. For an example of how the process of developing a procedure for virtual development is realized see Becker et al. (2005). Procedures do not, however, fully determine the recurrent behaviour patterns. And the procedures are only one amongst other influences on recurrent behaviour patterns. In the present context, however, they are an important influence. In what follows, when we mention ‘technology’ we consider this to comprise both its physical and social components (such as stable practices). Under the assumption that surprising results are a fixed proportion of the number of results (runs) carried out. Our focus in this chapter is on exploring the organizational conditions that have an impact on whether agents in the organization explore or exploit. We take agents’ dispositions to explore as given, and focus on what diﬀerence various organizational environments make for whether agents in that organizational setting will explore.

244

Organizational routines and organizational change and innovation

BIBLIOGRAPHY Adler, P.S. (2007), ‘The bureaucratization of innovation: a Marxist analysis’, mimeo, University of Southern California. Adler, P.S. and B. Borys (1996), ‘Two types of bureaucracy: enabling and coercive’, Administrative Science Quarterly, 41(1), 61–89. Aiken, M., S.B. Bacharach and J.L. French (1980), ‘Organizational structure, work process, and proposal making in administrative bureaucracies’, Academy of Management Journal, 23, 631–52. Anderson, Neil and Rosina M. Gasteiger (2007), ‘Helping creativity and innovation thrive in organizations: functional and dysfunctional perspectives’, in Janice Langan-Fox, Cary L. Cooper and Richard J. Klimoski (eds), Research Companion to the Dysfunctional Workplace, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 422–36. Andrews, J. and D.C. Smith (1996), ‘In search of the marketing imagination: factors aﬀecting the creativity of marketing programs for mature products’, Journal of Marketing Research, 33, 174–87. Becker, Markus C. (2004), ‘Organizational routines: a review of the literature’, Industrial and Corporate Change, 13(4), 643–77. Becker, Markus C. and Francesco Zirpoli (2006), ‘Problem-solving in new product development’, Logic Journal of the IGPL [Interest Group in Pure and Applied Logics], 14(2), Special Issue on ‘Abduction, practical reasoning and creative inferences in science’, pp. 379–90. Becker, Markus C., Thorbjørn Knudsen and James G. March (2006), ‘Schumpeter, Winter, and the sources of novelty’, Industrial and Corporate Change, 15(2), 353–71. Becker, Markus C., Pasquale Salvatore and Francesco Zirpoli (2005), ‘The impact of virtual simulation tools on problem-solving and new product development organization’, Research Policy, 34(9), 1305–21. Benghozi, Pierre-Jean (1990), ‘Managing innovation: from ad hoc to routine in French telecom’, Organization Studies, 11(4), 531–54. Bohn, Roger E. (1994), ‘Measuring and managing technological knowledge’, Sloan Management Review, 36(1), 61–73. Burns, T. and G.M. Stalker (1961), The Management of Innovation, London, UK: Tavistock. Cardinal, L.B. (2001), ‘Technological innovation in the pharmaceutical industry: the use of organizational control in managing research and development’, Organization Science, 12, 19–36. Cohen, Michael D., Roger Burkhart, Giovanni Dosi, Massimo Egidi, Luigi Marengo, Massimo Warglien and Sidney G. Winter (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5(3), 653–98. Cooper, Robert G. (1990), ‘Stage-gate systems: a new tool for managing new products’, Business Horizons, 33(3), 44–54. Cooper, Robert G. (1999), ‘From experience – the invisible success factors in product innovation’, Journal of Product Innovation Management, 16(2), 115–33. Cooper, Robert G. and Elko J. Kleinschmidt (1995), ‘Benchmarking the firm’s critical success factors in new product development’, Journal of Product Innovation Management, 12(5), 374–91.

The role of procedures and stable behaviour patterns in innovation

245

Cyert, Richard M. and James G. March (1963/1992), A Behavioral Theory of the Firm, 2nd edn, Oxford, UK: Blackwell. Damanpour, Fariborz (1991), ‘Organizational innovation’, Academy of Management Journal, 34(3), 555–91. Davila, Tony (2000), ‘An empirical study on the drivers of management control systems’ design in new product development’, Accounting, Organizations and Society, 25(4–5), 383–409. De Jong, J.P.J. and R. Kemp (2003), ‘Determinants of co-workers innovative behaviour: an investigation into knowledge intensive services’, International Journal of Innovation Management, 7, 189–212. Dosi, G., R.R. Nelson and S.G. Winter (2000), ‘Introduction: the nature and dynamics of organisational capabilities’, in G. Dosi, R.R. Nelson and S.G. Winter (eds), The Nature and Dynamics of Organisational Capabilities, Oxford, UK: Oxford University Press, pp. 1–22. Eisenhardt, Kathleen (1989), ‘Building theories from case study research’, Academy of Management Review, 14(4), 532–50. Feldman, Martha S. and Brian T. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48(1), 94–118. Fleming, Lee and Olav Sorenson (2004), ‘Science as a map in technological search’, Strategic Management Journal, 25, 909–28. Ford, C.M. and D.A. Gioia (2000), ‘Factors influencing creativity in the domain of managerial decision making’, Journal of Management, 26, 705–32. Gilson, L.L., J.E. Mathieu, C.E. Shalley and T.M. Ruddy (in press), ‘Creativity and standardization: complementary or conflicting drivers of team eﬀectiveness’, Academy of Management Journal, forthcoming. Greve, Henrich (2007), ‘Exploration and exploitation in product innovation’, Industrial and Corporate Change, 16, 945–75. Griﬃn, A. and J.R. Hauser (1996), ‘Integrating R&D and marketing: a review and analysis of the literature’, Journal of Product Innovation Management, 13, 191–215. Gupta, Anil K., Ken G. Smith and Christina E. Shalley (2006), ‘The interplay between exploration and exploitation’, Academy of Management Journal, 49(4), 693–706. Hargadon, Andrew and Robert I. Sutton (1997), ‘Technology brokering and innovation in a product development firm’, Administrative Science Quarterly, 42(4), 716–49. Hargadon, Andrew and Robert I. Sutton (2000), ‘Building an innovation factory’, Harvard Business Review, May–June, 157–66. Hodgson, Geoﬀrey M. (2008), ‘The concept of a routine’, in Markus C. Becker (ed.), Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 15–28. Jelinek, Marianne and Claudia B. Schoonhoven (1993), The Innovation Marathon, San Francisco, CA, USA: Jossey-Bass. Johnson, J.D., M.E. Meyer, J.M. Berkowitz, C.T. Ethington and V.D. Miller (1997), ‘Testing two contrasting structural models of innovativeness in a contractual network’, Human Communication Research, 24, 320–48. Kahn, Kenneth B., Gloria Barczak and Roberta Moss (2006), ‘Dialogue on best practices in new product development. Perspective: establishing an NPD best practices framework’, Journal of Product Innovation Management, 23, 106–16.

246

Organizational routines and organizational change and innovation

Katila, R. and G. Ahuja (2002), ‘Something old, something new: a longitudinal study of search behavior and new product introduction’, Academy of Management Journal, 45(6), 1183–94. Knudsen, Thorbjørn (2008), ‘Organizational routines in evolutionary theory’, in Markus C. Becker (ed.), Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 125–51. Levinthal, Daniel (1997), ‘Adaptation on rugged landscapes’, Management Science, 43(7), 934–50. Levinthal, Daniel and Claus Rerup (2006), ‘Crossing an apparent chasm: bridging mindful and less-mindful perspectives on organizational learning’, Organization Science, 17(4), 502–13. March, James G. (1994), A Primer on Decision-Making, New York, USA: Free Press. March, James G. and Herbert Simon (1958/1993), Organizations, Oxford, UK: Blackwell. Mellor, S. and J.E. Mathieu (1999), ‘A discriminant validity study of aggregatelevel constructs and measures of local union formalization, centralization, and innovation’, The Journal of Psychology, 133, 669–83. Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Belknap Press of Harvard University Press. Ocasio, William (1997), ‘Towards an attention-based view of the firm’, Strategic Management Journal, 18, 187–206. Ohly, Sandra, Sabine Sonnentag and Franziska Pluntke (2006), ‘Routinization, work characteristics and their relationships with creative and proactive behaviors’, Journal of Organizational Behavior, 27(3), 257–79. O’Reilly, C. and M.L. Tushman (2004), ‘The ambidextrous organisation’, Harvard Business Review, 82(4), 74–81. Pavitt, Keith (2002), ‘Innovating routines in the business firm: what corporate tasks should they be accomplishing?’, Industrial and Corporate Change, 11(1), 117–33. Pentland, Brian T. and Henry Rueter (1994), ‘Organizational routines as grammars of action’, Administrative Science Quarterly, 39, 484–510. Pettigrew, Andrew M. (1990), ‘Longitudinal field research on change: theory and practice’, Organization Science, 1(3), 267–92. Pierce, J.L. and A.L. Delbecq (1977), ‘Organizational structure, individual attitudes and innovation’, Academy of Management Review, 2, 27–37. Schön, Donald A. (1983), The Reflective Practitioner, New York, USA: Basic Books. Schulz, Martin (2008), ‘Staying on track – a voyage to the internal mechanisms of routine reproduction’, in Markus C. Becker (ed.), Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 228–55. Schumpeter, Joseph A. (1934), The Theory of Economic Development: An Inquiry into Profits, Capital, Interest, and the Business Cycle, Cambridge, MA, USA: Harvard University Press. Schumpeter, Joseph A. (1942), Capitalism, Socialism and Democracy, New York, USA: Harper Torchbooks. Scott, S.G. and R.A. Bruce (1994), ‘Determinants of innovative behavior: a path model of individual innovation in the workplace’, Academy of Management Journal, 37, 580–607.

The role of procedures and stable behaviour patterns in innovation

247

Thomke, S.H. (1998a), ‘Simulation, learning and R&D performance: evidence from automotive development’, Research Policy, 27, 55–74. Thomke, S.H. (1998b), ‘Managing experimentation in the design of new products’, Management Science, 33, 743–62. Thomke, S.H. and T. Fujimoto (2000), ‘The eﬀect of “front-loading” problemsolving on product development performance’, Journal of Product Innovation Management, 17, 128–42. Turner, Scott F. (2003), ‘The inertia of innovation: temporal routines for generational product innovation in computer software’, PhD Thesis, University of North Carolina, Chapel Hill, NC, USA. Turner, Scott F., Richard A. Bettis and Will Mitchell (2004), ‘Temporal routines for generational product innovation in computer software’, Working Paper, University of North Carolina, Chapel Hill, NC, USA, draft of 8 July 2004. Tushman, M.L. and C. O’Reilly (1996), ‘Ambidextrous organizations: managing evolutionary and revolutionary change’, California Management Review, 38(4), 8–30. West, M.A. (2002), ‘Sparkling fountains or stagnant ponds: an integrative model of creativity and innovation implementation in work groups’, Applied Psychology: An International Review, 51, 355–424. Wheelwright, S.C. and K.B. Clark (1992), Revolutionizing Product Development, New York, USA: Free Press. Winter, Sidney G. (1971), ‘Satisficing, selection, and the innovating remnant’, Quarterly Journal of Economics, 85(2), 237–61. Winter, Sidney G. (1984), ‘Schumpterian competition in alternative technological regimes’, Journal of Economic Behavior and Organization, 5, 287–320. Yin, Robert K. (1994), Case Study Research: Design and Methods, Thousand Oaks, CA, USA: Sage.

11.

The diﬃcult creation of novel routines: persistence of old habits and renewal of knowledge base in French SMEs Frédéric Huet and Nathalie Lazaric

1.

INTRODUCTION

The relation between institutions and individual behaviour has been widely debated in the old American institutionalism, according to which collective learning rests on individual habits, routines and other types of more or less formalized practices (Commons, 1934; Veblen, 1914). New interest in the notion of routine has recently arisen, particularly following Nelson and Winter’s work (1982), which highlighted the relative permanency of firms’ behaviours but also their capacity to innovate. Using a Schumpeterian framework, these authors free themselves from the traditional institutional framework and consider that processes of routine selection respond essentially to external regularities. Yet, a careful re-examination of the concepts of habits and routines shows the likeness of both notions, in terms of their properties and of their ability to integrate changes (Hodgson, 1993; Lazaric, 2000; Lorenz, 2000). The notion of routine is increasingly used to analyse microeconomic change (Feldman, 2000, 2004; Pentland and Feldman, 2005; Becker et al., 2005). In this respect, a re-examination of institutions’ role would enable us to better identify and understand the forces behind these changes, which are not exclusively related to cognitive contingencies (Nelson, 1994; Nelson and Sampat, 2001). The persistence of old habits and the diﬃculty of creating novel routines notably through cooperation gives us the opportunity to examine the organizational and institutional dynamics from an evolutionary perspective. Pioneering eﬀorts have already been made in this field to reach beyond the traditional Schumpeterian rhetoric that describes the process of change in terms of rupture, and to build an analytical framework that is able to take this emergence into account (Costello, 1996; Endres and Woods, 2006; Acs and Varga, 2005). We shall therefore 248

Persistence of old habits and renewal of knowledge base in French SMEs 249

attempt to continue in this endeavour while presenting the methodological problems we encountered. For this purpose, we shall use diverse sources of data (quantitative and qualitative) for tracking persistence of routines and their diﬃcult renewal in SMEs. This leads us to portray firms and their transactions via potential cooperation which may trigger (or not) some opportunities to explore new ways of doing things. Analyses of innovation in SMEs often emphasize that, in increasingly fast-changing and unpredictable environments, small firms frequently have insuﬃcient resources to undertake innovation projects and to renew their competencies (Liao et al., 2003; Nooteboom, 1994). In the face of these limitations, mobilizing external resources can be considered as an organizational strategy enabling firms to solve this internal dilemma. Cooperation can then be seen as a privileged means of access to innovation, in that it provides a way of exploring new paths to competency renewal. The exploration potential oﬀered by cooperation is undeniable but the question of the adaptation of this organizational form by small enterprises – whose innovation capacities are limited – is as yet unresolved (Davenport, 2005). We shall discuss these issues with data coming from an original survey in SMEs. These empirical results will give us the opportunity to debate changes and to observe the evolution of inertia inside current practices. Section 2 will introduce some new analytical framework for observing changes and regularities inside entrepreneurs’ cognitive frames and their institutional environment. Section 3 will present the data collected and some econometrics results in two models called ‘routine transactions’ and ‘strategic transactions’. Section 4 will provide some complementary description with further qualitative investigation in order to zoom in on some micro elements showing the diﬃculty of introducing organizational change. Section 5 will underline the question of change and its observation for researchers. We shall conclude by underlining the explanatory character of our research which still needs additional quantitative and qualitative data. Methodological issues and limitations will be presented.

2.

NEWNESS, COGNITIVE REGULARITIES AND LOCALIZED LEARNING: WHAT’S NEW FOR DEPICTING SMES

2.1

Images as Regularities and Frame for Entrepreneurs

SMEs are very often depicted in terms of flexibility and innovation while big companies may appear as a source of bureaucratization with a potential

250

Organizational routines and organizational change and innovation

organizational inertia (Schumpeter, 1942; Hannan and Freeman, 1984). In this chapter, we go beyond this idea, arguing that SMEs might be reluctant to introduce organizational change that would disrupt their habits and current ways of doing things. This leads us to consider institutional or cognitive inertia as a crucial obstacle to change. This is particularly significant for SMEs, which are characterized by more direct coordination mechanisms and fewer codification processes and procedures. Even if this reduced bureaucratization can explain more flexibility, it can also explain that competences for SMEs rely on a strong base of tacit knowledge and consequently can lead to a bigger persistence of existing cognitive frameworks and resistance to change (Hodgson, 1988; Nooteboom, 2000). For this reason, cooperation, which can aﬀord potential practice renewal being a source of innovation and organizational change, may be perceived as a disruptive factor that could be avoided. This fear of ‘newness’ which may appear, at first sight, as rather counterintuitive, takes root in the traditional SMEs’ vision of the active entrepreneurs who are surviving in a competitive environment, and is a belief largely diﬀused by Schumpeter some decades ago. But those entrepreneurs – as human beings – have their own frameworks, their own cognitive automatisms and their intrinsic motivation, leading them either to resist such organizational change or to be at the source of it. Cognitive frameworks result both from these internal processes and the local and cultural environment (Bandura, 1986; Witt, 2000). In other words, cognitive frameworks follow from the co-constitution of action and perception, as defended by the constructivist approach (see notably Weick, 1995, 2001, on this dimension). For Boulding (1956), images play the role of intermediation between the perception of raw data and the internal value system. Each human action is induced by man’s own images, but these images can be altered or revised by the actions themselves: Every man has a self-image, which includes a picture of his location in space, his acknowledgement of being part of a time flow, the perception of the universe around him as a world of regularities and the sensation of being part of a human relational network. He then has an image to evaluate reality (‘value image’) which intervenes in his relationship with the external environment, embedding information with meanings, an ‘aﬀectional or emotional image’ which provides him with feelings, attitudes and motivations, and a ‘public image’ which helps to compare his personal views with those shared collectively. (Boulding, 1956: 14)

In this perspective, images are a way to interpret information and to make sense of the environment. Images create temporarily stable cognitive frameworks bearing individual and collective regularities: ‘In the

Persistence of old habits and renewal of knowledge base in French SMEs 251

meanwhile, individual imagery has a relevant social function because it enables collective sharing of values and meanings. From this point of view the image has cohesive power which may exert a strategic function in organisational contexts and in co-operative interaction’ (Patalano, 2009). But these images are not inert, and they evolve thanks to the experiences that are likely to contradict the existing images: ‘As each event occurs, however, it alters my knowledge structure or my image. [. . .] Every time a message reaches him, his image is likely to be changed in some degree by it, and as his image is changed, his behaviour pattern will be changed likewise’ (Boulding, 1956: 5). This insight highlights that action might also modify cognition either deliberately or simply by accident (Weick, 2000; Greve, 1999). In other words these constitutive regularities are entangled and the interpretative frameworks not only guide performance but are shaped by it, as ‘ostensive routines’ are guided by ‘performative routines’ or might be constrained or triggered by them (Feldman, 2000; Pentland and Feldman, 2005). Innovation may, in this way, be created by an innovator who may modify the current thinking on the economic activity thanks to the emergence of less stereotyped images in some specific context (Patalano, 2008). Regularities also have their origin in ‘cognitive automatisms’, which are generated by a stabilization inside the ‘procedural knowledge’ that allows quicker memorizing when circumstances appear to be similar (Bargh, 1997; Cohen and Bacdayan, 1994). These potential automatisms, also taking their origin in the ‘declarative knowledge’ – that is the representational level – help human beings to find a predictable behaviour in a dynamic environment and to integrate some plasticity in the solving of new problems that the mind has not yet memorized (Lazaric, 2008, for a longer discussion on this point). Images are part of this system as they produce regularities inside the procedural knowledge as well as new insights in the declarative knowledge not always put in practice, that is transformed by the mind into a purposeful cognitive act. 2.2

Potential Automaticity and Discussion around the ‘Institutionalized Mind’

The interplay between the individual and the collective levels of action is far from being neutral for our purpose (Dopfer, 2007). Indeed, entrepreneurs shape their judgement, beliefs and acts not only by themselves but also in interactions with others. And the nature of these micro-interactions can produce ‘recurrent interacting patterns’ that need to be carefully observed (Cohen et al., 1996). In the old American institutionalism, Commons (1934, 1950) has proposed a taxonomy of interactions linked to the type of

252

Organizational routines and organizational change and innovation

knowledge involved (Dutraive, 2009). ‘Routine transactions’ are the ones related to habitual activities involving stabilized knowledge (embodied in rules), and ‘strategic transactions’ are those related to situations of novelty implying new practices and new opportunities and for which there is no stabilized knowledge and rules of thumb. In other words, ‘routine transactions’ are stabilized procedures deeply entrenched in the procedural memory of the entrepreneur, whereas ‘strategic transactions’ concern new ways of doing things that are not yet classified by the human mind. Indeed, for Commons, deliberation and calculative processes are not always mobilized and can consciously trigger past habits when they are appropriate. However, in some circumstances the mind may reveal ‘a creative agency looking towards the future and manipulating the external world and other people in view of expected consequences’ (Commons, 1934: 7; Hodgson, 1988). In short, an institution has to be understood as the working rules of collective action that may restrain individual deliberation and play a cognitive role by creating ‘institutionalized minds’ and ‘institutionalized personalities’ (Commons, 1934: 874). This entanglement between individual and collective actions is very clear: When a new worker goes into a factory or on a farm, or when a beginner starts in a profession or a business, everything may be novel and unexpected because not previously met in his experience. Gradually he learns the ways of doing things that are expected from him. They become familiar. He forgets that they were novel when he began. He is unable even to explain them to outsiders. They have become a routine, taken for granted. His mind is no longer called upon to think about them. We speak of such minds as institutionalized. But all minds are institutionalized by whatever habitual assumptions they have acquired and take for granted, so that they pay no attention to them except when some limiting factor emerges and goes contrary to what they are habitually expecting. (Commons, 1934: 697–8)

Commons – but also Veblen and the main authors belonging to the ‘old’ evolutionary thought – invites us to scrutinize both the mechanisms of change brought about by the individuals (the ‘upward causation’ having an impact on the organization) and the changes within the organization (the ‘reconstitutive downward causation’ that aﬀects the individuals) (Hodgson, 2007: 108)). Routines clearly lie between these two levels of analysis because they are enacted by individuals in a social context that can regulate the relative level of autonomy (Becker et al., 2005, see also Giddens, 1984). This interplay between the individual and collective dimensions has been very well illustrated in the literature in which entrepreneurs are not always capable of taking the so-called ‘best’ decision because of a vast amount of unreliable information. For this reason, they may have relevant heuristics

Persistence of old habits and renewal of knowledge base in French SMEs 253

describing similar contexts, which they use in order to analyse the competitive structure of the environment (Porac and Thomas, 1990). Their individual images are also framed by collective actions of the local environment. Entrepreneurs may act in a tied manner, not because they have their own cognitive limits but because the vast quantity of information they might collect is not always conducive to making the correct decision. This may push them to adopt mimetic behaviours dealing with the uncertainty they may have in making their judgement (Greve, 1998). Adopting local imitative behaviour means that in some circumstances the entrepreneur will fail to examine all the possible actions they could trigger (Kahneman, 2003). This myopia is explained by the voluntary ignorance of facts and data but also by the willingness to reduce learning and costs of searching for information (Kirzner, 1979). This localized learning – induced by various vicarious learning processes – is present not only on the industrial level but also on the local level (Maskell and Malmberg, 2007). This can produce a deliberate unwillingness to absorb new knowledge in order to avoid redefining the deeply entrenched procedural knowledge that matches the current vision. This willingness to stay on ‘routine transactions’ and to keep away from the creation of new ‘strategic transactions’ is illustrated by the famous exploration/exploitation dilemma (Levinthal and March, 1993; Greve, 2007). This compromise shows that exploitation not only increases the probability to perform organizational routines again but simultaneously avoids exploration by reducing the resources available for research. This issue is crucial for SMEs that could act in a local environment with reduced resources which would prevent them from fully investing in the renewal of routines. For this reason, the fear of ‘newness’ is not always a pure cognitive process but also a dynamic in which resources are linked to the size of the firm, leading it to overestimate the cost of organizational change induced by innovation. The ability to conduct cooperation is seen as an opportunity to reinforce existing capabilities and to produce innovation brought about by external factors (Cattani, 2005; Greve, 1998, 2007). This argument is rather well grounded for our purpose since cooperation not only introduces an organizational change inside current practices but also generates a confrontation of the ‘images’ shaping the way to do things. It creates a kind of altercation inside entrepreneurial frameworks, and thus can be avoided. Novelty may also induce new ‘combinatorics of routines’ because of a change in some micro-interactions which may generate a disruptive change at the meso level (Schumpeter, 1934; Winter, 1975; Becker et al., 2006). For this purpose, the organizational inertia should not be seen as a purely defensive attitude but also as a way to maintain endogenous changes in the micro-interaction, helping to contain them in some predictable limits.

254

Organizational routines and organizational change and innovation

Mindful reflexivity (Langer and Moldoveanu, 2000) and motivation about organizational change is thus necessary to overcome these obstacles but not always suﬃcient (Howard-Grenvillle, 2005). This implies that motivational factors inside current practices should be in accordance with the change introduced at a cognitive level. The perception and the image of change are crucial here and concern both the declarative and procedural form of knowledge, that is to say the representation of change and its eﬀective implementation. In this perspective, the change in routines should not be seen as a fateful coincidence related to external and disruptive factors, but as a crucial ingredient to the revitalization of individuals and organizations. This leads us to reconsider the very meaning of the term ‘routine’ and to focus on individual and collective memorization processes: ‘Routines now appear to be more mindful and more variable and to consume more attention than was first thought. The continuum of mindful action that had previously been masked has now become more diﬃcult to ignore’ (Weick and Sutcliﬀe, 2006: 522). In order to create a reflexivity on routines, mindfulness matters and is materialized by an intention and a capacity to absorb change both at the motivational and cognitive levels (Huet and Lazaric, 2008; Lazaric et al., 2008). Mindful attitude may be defined, for our specific field, as a capacity to go beyond ‘routine transactions’ in order to change the procedural knowledge embedded in entrepreneurs’ minds and entrepreneurs’ ways of doing things. Mindful attitude is an explorative behaviour that has to be adopted to generate ‘strategic transactions’, that is transactions that are not always known in advance and which may trigger unpredictable change inside organizations. This change will be observed here with a methodology allowing us to zoom from the micro to the meso level (or the collective level).

3.

HOW TO OBSERVE THE CHARACTERISTICS AND THE NATURE OF INTERACTIONS IN SMEs

3.1

Presentation of the Data and Methodology

The sample of firms examined in our empirical study (small and medium enterprises in the metal industry) was not chosen randomly. Indeed, the dynamic of SMEs has seldom been studied with an evolutionary or institutional approach. Yet, as Brette (2003) acknowledges, it is essential to observe all aspects of the evolution of behaviours and habits within firms in order to understand what has caused them to change.

Persistence of old habits and renewal of knowledge base in French SMEs 255

E1

E2 E3

Figure 11.1

Presentation of the diﬀerent samples

The diﬀerent selected samples The analyses and results presented here are drawn from a survey conducted in 2003 and 2004 on cooperation in French SMEs. The cooperative practices of four sectors have been examined in this survey: the agro-food sector; the metalwork industry; the electronic component sector; and computer service firms.1 Our study is based on data corresponding to a sample of enterprises divided into four overlapping sub-samples. Figure 11.1 indicates the articulation between these diﬀerent sub-samples. The overall sample provides information about enterprises employing between 19 and 500 employees (which fits the common definition of SMEs). The first stage of the study is relative to the construction of a statistical database on French SMEs. The data was collected through telephone questionnaires2 administered between Spring 2001 and Spring 2002. Sample E1 comprises data on 637 French SMEs: this data concerns the generic characteristics of the firms surveyed. Sample E2 comprises the firms in sample E1 that claimed to have cooperative relations with other firms (relations of diﬀerent types: client supplier, relations between competitors, Economic Interest Group. . .), in addition to commercial relations. This sample of 97 firms contains information on what the interviewees considered as their main cooperative relationship: characteristics of the relationship and of the partner, evolution of the relationship, impact of the cooperation on the firm’s activities, flow of information between the partners. The number of firms contained in sample E2 indicates that only 15 per cent of all the firms surveyed declared that they had cooperative relations with other firms. Because of the bias associated with conducting the interviews by telephone, this cooperation rate cannot be considered as representative of the cooperation rate of French SMEs.3 However, it confirms the low propensity of these small enterprises

256

Organizational routines and organizational change and innovation

to cooperate with other firms. Partnership remains a relatively marginal feature of SMEs and consequently must be given particular attention if one is to understand the cooperative dynamics that develops among those firms. Sample E3 is relative to the second, qualitative, part of the study. Once the above-mentioned statistical data collection was completed, a number of firms were selected from sample E2 so that further interviews could be conducted with firms involved in cooperative relationships to complement the information provided by the database. Thus, 21 interviews were conducted and transcribed and were aimed at gaining a more detailed understanding, through concrete examples, of cooperative practices among SMEs (see Appendix 11.1, Table A11.4). These interviews provided us with qualitative information that could then be compared with the statistical results. The results of our study have enabled us to build indicators translating the previously described analytical elements, notably to analyse both strategic and cognitive dimensions of transactions. In addition to these factors of learning, indicators of cooperation performance have been constructed. For each category of learning factors, three dummy variables have been defined. The items of these indicators take value 1 if, according to our hypotheses, they have a positive eﬀect on learning in cooperative relationships. Thus the econometric model includes 12 independent variables. Table 11.1 summarizes these items. The interaction capacity is measured through the degree of complementarity of the competencies mobilized in the cooperation. The competencies that are considered diﬀerent and complementary imply a cognitive distance conducive to learning, whereas competencies that are considered similar reduce learning opportunities. Furthermore, being involved in activities of technological co-development requires the mobilization of competencies that can generate learning-conducive synergies. Finally, the size of the firms accounts for the cognitive distance in its more organizational dimension, namely the compatibility of the partner firms’ modes of organization and management. The intentions of interaction are first of all measured via the presence of collective work teams that are likely to promote the exchange of knowledge, and in particular of tacit knowledge (Beise-Zee and Rammer, 2006). The type of information exchanged between the partners is also indicative of their intentions in terms of cooperating. Exchanges of information considered sensitive or strategic are indicative of strong intentions of interactions. Finally the age of the relationship (when the latter is over two years old) indicates an intention to continue and further develop interactions with the partner.

Persistence of old habits and renewal of knowledge base in French SMEs 257

Table 11.1

Indicators of learning factors

Analytical category Indicator

Variables

Interaction capacity Technological cooperation (COOPTECH) Complementarity of competencies (CAMO) Size of the partner firm (TAPA) Intentions of Direct interactions interaction (EQTR) Circulation of information (ECIN)

✓

Duration of the relationship (DUREL) Absorptive capacity R&D activities (ACRD) Innovating firm (INNOV) New activities (MODAC) Intentions of Cooperative absorption behaviour (PROAC)

✓

Initial innovation goals (OBINNO)

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

Commitment to cooperation (EGT)

✓

✓

Activities of co-development (51) Commercial activities only (50) Diﬀerent competencies (51) Similar competencies (50) Similar (50) Diﬀerent (51) Common working teams (51) No common working teams (50) Exchange of strategic information (51) Exchange of non strategic information (50) Over 2 years (51) Less than 2 years (50) Formal R&D (51) No R&D (50) Recent innovation (51) No recent innovation (50) Recent development (51) No recent new activities (50) Proactive, initiated the partnership (51) Reactive, talked into becoming a partner (50) Initial definition of initial innovation goals (51) No innovation goals defined initially (50) Tangible signs of commitment (active participation, specific investments) (51) No tangible signs of commitment (50)

Some of the indicators of absorptive capacity used here have already been used in previous studies (see Cohen and Levinthal, 1990). They are the development of innovation (product, service or process innovation) and the presence of R&D activities. In addition to those, we have used an

258

Organizational routines and organizational change and innovation

indicator relative to reorganizations and recent developments that have taken place in the firm. Thus, these three indicators reflect a certain evolution in the firm’s knowledge base, and therefore a capacity to learn and assimilate new knowledge. The last category of indicators is related to the intention of absorption. We measure the latter via the firm’s proactiveness towards cooperation. Proactiveness on the firm’s part (for example when the firm actively seeks partners, and initiates relationships) is indicative of expectations and intentions that are more conducive to learning than a reactive behaviour in which the firm is more or less talked into or even forced into the cooperative relationship. Similarly, the existence of tangible signs of commitment to the relationship indicates that the firm has a strong intention to develop a cooperative relationship (Fukugawa, 2006). Finally, initiating a relationship with objectives in terms of innovation, learning and knowledge development indicates that the firm has strong intentions regarding absorption. The question of measuring the performance of the cooperation raises diﬃculties at two levels: that of the overall performance of the cooperation, and, more relevant to our study, that of the measurement of the learning achieved through the cooperation. With regard to the measurement of the performance of the cooperation, one first diﬃculty is that of identifying objective indicators: indeed it is diﬃcult to distinguish clearly the outcomes of the cooperation itself from the results of the internal activities of the firm. Furthermore, how can the outcomes of the cooperation be compared to alternative solutions that were not selected? Faced with these methodological diﬃculties in evaluating the performance of the cooperation on the basis of objective indicators, we opted to first conduct a subjective evaluation. Then, in order to overcome the problem presented by the small size of our sample (97 firms), we chose to construct dummy variables for the measurement of the exploration (for learning) and the exploitation.4 Furthermore, this rather subjective approach – which is based on an evaluation by the interviewee him/herself – is all the more necessary as the performance of the cooperation in terms of learning is diﬃcult to quantify objectively. How does one evaluate the renewal and evolution of organizational knowledge bases? Here again, the diﬃculty lies in diﬀerentiating the performances in terms of learning from other performances that might be unrelated to the former. Indeed, learning is not the only purpose of cooperation, and the performance of the latter therefore cannot be evaluated by exclusively taking into account its outcomes in terms of learning. In order to respond to the need to identify the performances in terms of learning and the factors that aﬀect the intensity of the learning process, we have distinguished two categories of indicators.

Persistence of old habits and renewal of knowledge base in French SMEs 259

Table 11.2

Indicators of the nature of the transaction

Indicators

Scales: Impact of the relationship on . . .

Strategic transactions

➤ ➤ ➤

Routine transaction

➤ ➤ ➤

The development of new innovation capacities Quality improvement Modification of work practices Productivity and costs Commercial positioning The financial situation of the firm

First of all, ‘strategic transactions’ indicators – which reflect the outcomes of the cooperation that imply a renewal of the knowledge bases – have been identified. The ‘exploratory’ nature of cooperative activities must be considered in the context of SMEs. Indeed, the ‘improvement of quality’ indicator may not at first sight seem to belong in the ‘strategic transactions’ category, and must be considered in the context of small firms, as the introduction of new organizational practices may, in SMEs more than in large firms, consist in a complete modification of existing practices.5 A second category of indicators is related to more static performances relative to the exploitation of available knowledge: ‘routine transactions’. The hypothesis underlying the decision to distinguish strategic transactions from routine transactions is that the factors of learning mentioned above aﬀect and discriminate the indicators of exploration rather than the indicators of exploitation. Table 11.2 summarizes the diﬀerent scales of evaluation of the performances of the cooperation. 3.2

Discussion about our Quantitative Results and Analyses

Econometric results The influence of the factors of learning on the performances of cooperation has been evaluated using two logit models. The latter enable us to run a regression of the ‘strategic transactions’ and ‘routine transactions’ indicators on these diﬀerent factors. The results of both models are presented in Table 11.3. Before proceeding to an in-depth discussion of these results, the heterogeneity of the two models must be noted. The independent variables we have used are more discriminant in the strategic transactions model than in the routine transactions model and therefore confirm our hypothesis that the factors aﬀecting organizational renewal diﬀer from those aﬀecting exploitation of existing practices. On the basis of the results obtained, we can now discuss not only the

260

Organizational routines and organizational change and innovation

Table 11.3 Categories of indicators

Logit regression on strategic and routine transactions models Variables

Strategic transactions model

Routine transactions model

Coeﬃcients (standard error) Constant Interaction capacity Intentions of interaction Absorptive capacity Intentions of absorption

COOTECH CAMO TAPA EQTR ECIN DUREL ACRD INNOV MODAC PROAC OBINNO EGT

2.75 (0.87) 1.20** (0.61) 21.31* (0.71) 0.56 (0.64) 0.34 (0.60) 1.15* (0.65) 0.10 (0.59) 20.88 (0.62) 1.43 **(0.61) 0.62 (0.59) 1.21** (0.60) 0.57 (0.68) 1.33** (0.63)

2.01 (0.77) 0.40 (0.55) 0.77 (0.63) 20.01 (0.60) 20.23 (0.57) 1.93 ***(0.59) 0.22 (0.55) 20.58 (0.59) 0.02 (0.56) 0.26 (0.60) 0.28 (0.56) 0.28 (0.65) 0.97 *(0.56)

Notes: * Significant at the threshold of 1%. ** Significant at the threshold of 5%. *** Significant at the threshold of 10%.

conditions and factors that are conducive to learning in the context of cooperation but also the validity and relevance of those indicators. The interviews we conducted (sample E3) will support and enrich this discussion. Initial conditions, path dependency and absorptive capacity With regard to the indicators of absorptive capacity, that related to innovation is the only one that has a discriminant eﬀect on learning processes. The other two, R&D activities and reorganization/modification of activities, are not. This heterogeneity indicates some diﬃculty in identifying – in particular with the indicators usually used for analysing large firms – the activities through which the absorptive capacity of SMEs develops. Indeed, reorganizations of activities might imply diﬀerent measures depending on the firms interviewed, and those measures might not necessarily contribute to the development of the firms’ absorptive capacity. As for R&D activities, they are relatively diﬃcult to evaluate in SMEs because they are, for the

Persistence of old habits and renewal of knowledge base in French SMEs 261

most part, conducted informally. Thus, besides the firms that declare that they perform formal R&D activities (that is which appear in their balance sheets), many others are involved in such activities, but in a much less visible fashion (Pacitto and Tordjman, 1999; Santarelli and Sterlacchini, 1990). Beyond the diﬃculty in evaluating the absorptive capacity of SMEs, the innovation indicator has a significant statistical influence on cooperationbased learning. The accumulation of knowledge generated through these innovative activities also implies a development of learning capacities that can be exploited in the framework of cooperations. For SMEs, possessing a capacity of absorption is therefore a factor that intensifies external learning opportunities. This role of a firm’s absorptive capacity therefore implies that initiating cooperation as a means of increasing the organization’s external learning capacities will be all the more successful if the firm itself is characterized by cognitive dynamism. In other words, the processes of external learning cannot substitute the internal processes but must, on the contrary, be envisaged as part of a dynamic of co-construction and self-reinforcement. Intention of interaction and circulation of information Among the four categories of learning factors identified, the intention of interaction is the one that seems to have the least influence on the firm’s ability to develop new competencies. Indeed the only indicator aﬀecting the intensity of these learning processes is that which is related to information exchange. It is, however, important to note that this increased exchange of information (considered strategic and sensitive) is a variable that also aﬀects the firm’s performances in terms of exploitation of existing knowledge. In other words, more than a condition for learning, the circulation of information appears to be a condition for cooperation. Our models show that the age of the relationship does not have any significant impact. However, we believe that this result does not so much indicate that learning only occurs at the beginning of the relationship, but rather that it is diﬃcult to state exactly when the relationship started. Indeed, the results appended to those presented here show that cooperative relationships are frequently a continuation of commercial or informal relationships. Thus, on the one hand, it is diﬃcult to state or determine the exact date on which the cooperation actually started, and on the other, the relationships that existed prior to the actual cooperation provided a favourable environment to the detection of learning opportunities, the realization of which the cooperation triggered and accelerated (Huet and Lazaric, 2004, 2008). Furthermore, the creation of common work teams has no statistical

262

Organizational routines and organizational change and innovation

impact on cooperation-based learning, which would tend to invalidate the hypothesis according to which interaction plays a central role in the learning processes necessary to the transfer and appropriation of tacit knowledge. The exchange of information, and therefore clarification and codification, is a necessary condition for cooperation; direct interactions are less crucial. However, an analysis performed specifically on the basis of the scale evaluating the impact of cooperation on work practices shows that this indicator of direct interactions becomes highly discriminant (Huet, 2004). Thus, when learning processes involve largely tacit and embedded knowledge, direct interactions prove crucial. The following quotation drawn from an interview with an entrepreneur of the metal sector, concerning the latter’s relationship with a client, is a good illustration of the role played by direct interactions in the renewal of work practices and particularly of how these interactions facilitate the coordination between a supplier and his or her client: ‘You [the employer] send them [the employees] there [to the clients], and when they come back they are the ones who tell you “be careful, don’t do that, there they don’t. . . .” You know, they often say that seeing is believing. . . well, it’s true’ (TM2 interview). In other words, our composite indicator ‘strategic transaction’ does not diﬀerentiate between the diﬀerent types of knowledge involved in the learning process, particularly in relation to their codified/tacit nature. An analysis based on a finer categorization of this knowledge (and produced) would probably reveal the role of the diﬀerent types of interaction according to the learning situations – including, but not exclusively, information exchange. Familiarity with the partners as a precondition for absorbing new knowledge Among the indicators of interaction capacity, seen above, two are discriminant. The size indicator, relative to the organizational cognitive distance, has no eﬀect on the learning process. Inversely, the presence of technological co-development activities influences the learning process. This distinction seems to show that, in the case of SMEs, the technological distance is more mobilized than the organizational distance. Thus, the learning process seems to be aimed at acquiring knowledge that is associated with the technical rather than the organizational competencies of each partner. The most significant result, in terms of interaction capacity, is related to the indicator of complementarity between the partners’ competencies. This indicator is, indeed, a discriminant indicator, but its negative sign indicates that the learning intensity is higher when there is a certain degree of similarity between the partners’ competencies. This result prompts us to examine the eﬀects of cognitive distance on the

Persistence of old habits and renewal of knowledge base in French SMEs 263

learning process. It tends to show that familiarity with the partner’s technological knowledge is a necessary condition for its absorption. Indeed, the more similar the partner’s knowledge base is to the knowledge initially possessed by the firm, the easier it is to combine and assimilate because it does not disrupt the existing managerial representations. But this potential aversion to exploration might lead the firm into competency traps caused by too great a proximity between the knowledge bases (Ahuja and Lampert, 2001).6 In our study, this result shows that learning opportunities arise when the cognitive distance between the partners is small. The following quotation drawn from an interview with an entrepreneur of the metalwork sector illustrates that without suﬃcient proximity between the two knowledge bases, the partners find it diﬃcult to identify learning opportunities: ‘When we visit the factory, we look around the workshop thinking: it’s nice, it’s fun, it’s noisy and it works well. We’re kind of curious about it but apart from that we’re not really interested’ (TM1 interview). Yet, the competencies of each partner are relatively accessible to the other partner, since they are all related to metal surface treatment activities. Our results do not deny that learning is possible, but they show that the learning process occurs through the combination of relatively similar knowledge bases. The tangible consequence of this characteristic, for SMEs, is that these learning experiences seldom generate what one would call ‘revolutionary’ knowledge or radical transformations of existing knowledge bases. This tendency is confirmed by the interviews we conducted, as indeed, out of the 21 firms we interviewed, only one produced, through cooperation with another firm, novel knowledge and technology (in the agro-food sector, innovation in the packaging of liquid products used in outer space missions). Finally, the analysis of the interaction capacity of the SMEs in our sample shows that cooperation as a means of creating learning opportunities is under-exploited. Because learning opportunities only emerge in SMEs if the cognitive distance between the partners is small enough, these opportunities represent relatively unsatisfactory solutions to the phenomenon of internal myopia, which raises the question of creation of novel routines or, at least, the potential introduction of new knowledge for making them change. Learning from cooperation and renewal of the knowledge base: a product of chance and deliberation This last category of indicators is related to the intentions of the firm with regards to the relationship and to the management of the relationship. A joint analysis of the three indicators enables us to identify and describe more precisely the firm’s objectives in cooperating with another firm and the nature of its commitment to the relationship.

264

Organizational routines and organizational change and innovation

Indeed, the two discriminant indicators are those that we could qualify as ‘generic’ in relation to the intended purposes of the cooperation: thus, the proactive cooperative behaviour and the tangible signs of commitment impact on the learning processes. Inversely, the indicator related to the objectives of innovation is not discriminant. In other words, cooperationbased learning requires proactiveness and strong involvement on the part of the firm, but does not necessarily require that its purposes be precisely or initially defined. The fact that the purposes of the cooperation are not precisely determined initially is well illustrated by the following quotation in which an entrepreneur of the electronic component sector expresses what he hopes to achieve by cooperating with a competitor: ‘We’re hoping that this will give us access to new opportunities’ (CE2 interview). Thus, expectations are not clearly defined initially; cooperation develops more on the basis of intentions to detect opportunities than on rational and predetermined strategies. Thus, cooperation seems to be a process through which learning opportunities are progressively unveiled rather than a clearly defined project with specific objectives. For example, in the context of a partnership between two agro-food firms, the aim of the two partners was to co-develop and co-produce so as to cope with the fluctuations in the demand. But the relationship also benefited from the joint exploitation of a patent registered by one of the partners: ‘I used to use a rigid tube that was on the market at the time; but it gave me problems. I benefited from his [the partner’s] patent’ (AA2 interview). This analysis of the intentions of absorption shows that the benefits of cooperation lie in the partners’ ability to broaden the potential scope of the cooperation and to be proactive enough to reveal new ‘areas of intervention’ to enrich the relationship. More than in an ex ante planning and identification of the content of the cooperative relationship, the intentions of absorption lie in the partners’ ability to facilitate the relational dynamic and to unveil new opportunities. Cooperation therefore requires new ways of doing things and new cognitive frames so as to avoid restricting the relationship to an overly constraining framework and so as to allow for the emergence of new opportunities and the creation of new transactions. In sum ‘strategic transactions’ and explicit exploration of opportunities are rarely fully mobilized ex ante but emerged for progressively changing existing practices and gradually transforming their inherent quality.

Persistence of old habits and renewal of knowledge base in French SMEs 265

4.

LOCAL ‘ATMOSPHÈRE’ AND GRADUAL CHANGE IN ENTREPRENEURIAL VISIONS: ZOOMING INSIDE THE METAL INDUSTRY

Zooming inside the metal industy where the rate of cooperation is lower is important to explain the diﬃculty of changing ways of doing things.7 Organizational routines prevailing in this sector therefore seem to be practices of non-cooperation. And it is important to understand how these transactions will evolve through rhetoric on cooperation and its implementation. Change in routines is a crucial phenomenon that we must measure with care, in order to go beyond the level of managerial rhetoric and its principles of action. Furthermore, the metal industry has undergone profound changes imposed by prime contractors in the aeronautics sector with the emergence of new shareholders and the implementation of new governance modes. Indeed the latter have forced subcontractors to implement new organizational methods so as to be able to provide integrated technological products rather than isolated products. Cooperation is one of the organizational changes promoted by contractors. Interviews we have conducted show that entrepreneurs are often illequipped to face these new requirements. The skills needed to respond to these demands do not merely consist of the technical skills around which these SMEs developed initially. The organizational (and inter-organizational) changes and the development of cooperative services required seem diﬃcult to integrate into the strategic positioning of these small firms. Thus studying these SMEs may give the opportunity to observe this change in their representations and their current transactions. In other words, interviewing SMEs’ managers has enabled us to jointly examine entrepreneurs’ visions, through their external manifestation with discourses and the firms’ actual experiences and activities. Considering principles and practices in a non-separate manner makes it possible to understand the representation/action combination underlying SMEs’ organizational dynamic, and to try to describe these interactions between both levels. 4.1

Emergence of New Frames Disrupting Existing Cognitive Regularities

The principles and frames that prevailed until recently, show unequivocally that cooperation used to be absent and was diﬃcult to envisage in the region studied: ‘There is hardly any cooperation; I’d say it is the law of the jungle’; ‘people are closed oﬀ’; ‘we are in a very individualistic world’. SMEs mostly operate in isolation; closed-door strategies prevail and interactions between firms are low despite their local concentration and

266

Organizational routines and organizational change and innovation

their quasi similar position towards contractors. The organizational stability rests on the compatibility and the coherence between individualistic representations or mental patterns and autarkic habits. Evolutions in subcontracting conditions and in the above-mentioned competitive environment force SMEs to implement organizational change: ‘We couldn’t function like that any more’. But changing ‘performative routines’ is a diﬃcult exercise in that it is dependent upon evolutions in representations: ‘Changing mentalities is the first diﬃcult stage. We need to take measures to coordinate the members’. Entrepreneurial visions both guide and are guided by current transactions. Whether these practices or experience feedbacks come from the firm itself or from actors of its environment, they make it possible to modify initial beliefs: ‘It’s easier because people are starting to know about the groups’; ‘They are ready to come together. They just hadn’t thought of it. They aren’t used to this.’ In other words, representations of cooperation are influenced by prior transactions. But representations evolve progressively according to experience and may lead the organization toward the adoption of some organizational change (via the introduction of new micro practices). The transitions from ‘individualistic’ beliefs to more ‘altruistic’ beliefs and from ‘autarkic’ practices to more ‘collective’ practices foster each other. We find here a first manifestation of the principle of cumulative causality at organization level, and therefore a coherence between its representations and actions. 4.2

Cumulative Change in Existing ‘Transaction Routines’ Introduced by Local Public Institutions

Two key actors in the region studied initiated the setting up of a consortium of companies with the purpose of grouping together their commercial activities: first, a few entrepreneurs, needing to find solutions to problems they were facing, had a pioneering role in the implementation of these cooperative initiatives: ‘We had to find a solution other than the natural solution’; ‘I see that the dynamic firms are those that are open’. Thus, new principles emerge without them necessarily translating immediately into new organizational routines: ‘When we thought of doing it, we didn’t know how to go about it’. The absence of judicial and administrative competencies initially limited the ‘concretization’ of these new entrepreneurial visions. The public institutions, the second key actor in the process, facilitated the implementation of these commercial consortiums and played the role of mediators: ‘I think that if he [the head of the local Chamber of Commerce and Industry] hadn’t helped us initially, we wouldn’t have been able to go all the way’; ‘and then thinking it was a good idea, he pushed the

Persistence of old habits and renewal of knowledge base in French SMEs 267

firms to come together’. The person in question confirms his support of and involvement in the development of such relations: ‘I think it is a concept that is particularly well adapted to the local environment’; ‘I pushed them; one of my roles is to “sell” the idea of collective eﬀort, and to tell them that their future depends on collaboration’. This articulation between the contributions of ‘pioneering’ entrepreneurs and of public institutions reveals the dynamic of change. Changes in the institutional environment (in our case illustrated by the local development policies) are inspired by renewed entrepreneurial visions, but they also lead to and catalyse the development of novel routines. Similarly, organizational changes would not have emerged without the contributions of this environment, for example the provision of judicial competencies to the SMEs. It appears that the institutionalization of cooperation does not consist of a linear dynamic – be it a top-down or a bottom-up dynamic – and that it rests on a principle of co-evolution characterized by multiple and cumulative causalities. The evolution is characterized by the ‘mutual mobilization’ of organizations and of their environment, which progressively leads to the institutionalization of cooperative relationships: ‘We don’t have that war any more. We even trust each other’. The emergence of new institutional practices has incrementally led to the routinization of cooperative activities and behaviours, not only among the firms but also in the region as a whole. As commented by an entrepreneur who recently set up a cooperative with a client, ‘They don’t need any explanation any more when they see the order form; let’s say they know exactly what to do. It just happens [. . .] it’s routine. . . so there’s no need to interfere’.

5.

DISCUSSION: SOURCES OF CHANGE AND STABILITY INSIDE ROUTINES

Change in routines has long been considered more likely to be determined by change in the environment (Cohen et al., 1996: 683). This criticism also applied, more widely, to the evolutionary theory of change (Vromen, 2006). Indeed Nelson and Winter’s works can, at first sight, be interpreted as meaning that the competitive environment and external forces are the only determinants of change in routines; but a more careful reading suggests otherwise. Indeed: Behavioural routines do change over time, in both desired and undesired ways. Such changes correspond to mutations in the biological theory, and without

268

Organizational routines and organizational change and innovation

them there could be no long-term evolutionary change. In so far as these changes are unintended and undirected, as many of them certainly are, the biological analogy is very close, and perhaps it can be extended to the realm of theoretical results. [. . .] Rather it is deliberate innovation, the product of directed eﬀort, typically undertaken in response to identifiable economic stimuli, and motivated by profit consideration. Thus, with a considered step away from the biological theory, it is proposed that deliberate innovations, large and small, be counted as the most significant subset of the changes in routine behaviours. (Winter, 1975: 102)

Thus, two sources of renewal are revealed in the Schumpeterian theory: on the one hand, the ‘combinatorics of routines’, which occurs through a combination of sub-elements (but though relevant, this approach is still in its infancy) (Becker et al., 2006: 362); and on the other hand, the unreliable process of replication of internal elements. Indeed a source of competitive advantage definitely lies in the firm’s ability to copy routines that exist in other firms; but also and above all in its ability to extend its own internal routines. This mechanism may be imperfect and costly, but it can also be a precious source of evolution and change. This reflection borrowed from Schumpeter and developed by Winter, tries to explain the extension and renewal of knowledge bases in a given competitive environment (Becker and Lazaric, 2003; Becker et al., 2006). In this regard, the setting up of micro practices in two Italian firms illustrates the limitations of this recombination of routines, which, in this specific case, is initiated and driven by the top management and which ‘operates more through repeated recombination patterns of stable factors than through a disruption of existing practices’ (Salvato, 2003: 83). This raises the ever-recurring question of novelty; in other words, how can novelty be generated out of existing elements (Aldrich, 1999; Demers, 2007)? This is the reason why an endogenous theory of change still needs to be explained in relation to the theory of routines in order to answer the thorny question of novelty, permanence and resistance to change. In short, more research in this direction is needed because: ‘Accounting for the reliable spread and diﬀerential adoption of routines is only part of the puzzle, however. What is missing is a theory of endogenous generation of distinctly novel routines’ (Becker et al., 2006: 361). Finally, it is also important to note that existing analyses of change in routines concentrate on internal or external changes, but few of them examine the possible interactions between both sources of change. Organizations have the ability to interpret external signs in many diﬀerent ways; as a result, the intentional or non-intentional nature of change is often diﬃcult, for an outsider, to observe.

Persistence of old habits and renewal of knowledge base in French SMEs 269

CONCLUSION The work and results presented in this chapter are in many respects exploratory. In order to reach a verdict and to give a more robust validation of the framework of analysis presented here, two types of developments can be envisaged. First of all, the sample on the basis of which we have built our econometric model is relatively small, and so, it would probably be useful to extend the analysis to a larger sample of firms (possibly with more sectors). Furthermore, this development would enable us to refine some of the indicators we have used here, particularly those related to the intentions of interaction. However, the studies that have been conducted up to now, though they have the merit of using larger, more representative samples, provide little information that would help analyse the factors of learning in cooperation, causing them to postulate, rather than question, the fact that cooperation automatically leads to learning. Beyond these methodological precautions, this study has helped us to understand better the nature and characteristics of organizational change in SMEs, notably via cooperation. First, the learning potential provided by cooperation seems often to be under-exploited in SMEs. While cooperation is now considered as an eﬀective means to innovation, this might be truer of large firms than SMEs. Furthermore, the strategy of innovation through cooperation cannot be substituted by the mobilization of internal learning processes. In other words, innovation strategies cannot be implemented by separating internal from external strategies, but must, on the contrary, be envisaged as an all-encompassing organizational strategy combining internal and external resources; thus some of the results obtained by other studies should be approached with caution (see Liao et al., 2003). Finally, the question of renewal of existing routines is all the more crucial as it shifts the focus from the planning of innovation projects to the active and continuous search for learning opportunities and the possible extension of the scope of the relationship. Additionally, with regard to firms’ absorptive capacity, our results highlight that the cooperative relationships developed by SMEs are often characterized by a small cognitive distance; this indicates a diﬃculty in renewing their existing transactions. Indeed, French SMEs are more inclined to enter into cooperation with other firms if they are already familiar with the activities of those firms. This explains their ability to rest on their current routine transactions and their diﬃculty in shifting from this position to generate some strategic transactions. Pre-existing learning and prior knowledge explains part of this story, but not all. Entrepreneurial frameworks and exisiting cognitive regularities explain another part, and

270

Organizational routines and organizational change and innovation

motivation is necessary to renew existing transactions and consequently organizational routines but – as we have shown in the metal industry – local atmosphère and the surrounding environment may impede real endogeneous change and prevent progression beyond individual and collective procedural knowledge. Routines may be path dependent if local entrepreneurs are not able to go beyond their prior cognitive regularities for creating new transactions that are not reduced to the product of the past and for generating newness in their mind and in their organization.

NOTES 1. These sectors have not been selected at random: one of them was selected because it presented the highest rate of cooperation (see CIS2 survey): electronic components; while another was chosen because its cooperation rate was far below average (metalwork). In the survey on innovating firms (CIS2), the electronic component sector has a cooperation rate of 49.2 per cent against 26.6 per cent in the metal sector; the average cooperation rate is 33 per cent (this percentage only concerns the industrial sector and does not take the size of firms into account). A service sector has been included here, as the CIS survey only considered industrial activities. 2. The decision to administer the questionnaire by telephone was made following a comparable survey conducted in Denmark (DISKO) and which had shown that the response to a telephone questionnaire was better than the response to posted written questionnaires. However, this method requires that the questionnaire be formulated as concisely as possible, the telephone interview lasting 20 minutes maximum. 3. Declaring that they did not have any cooperative relation with other firms enabled the interviewees to end the telephone interview at the end of the first part of the questionnaire. Thus, the cooperation rate observed with this method is lower than the real cooperation rate characterizing French SMEs as a whole. 4. These indicators were constructed on the basis of an average of three scales for each category of indicator (strategic transactions and routines transactions), and on a rating of 0 or 1. 5. For example, it has been shown that the implementation of the ISO 9002 standard in SMEs of the agro-food sector has led to a complete modification of cognitive automatisms and of organizational routines related to micro-tasks as well as to the diﬀerent departments of the firm, and to a new articulation of knowledge (Lazaric and Denis, 2005). 6. Indeed, these authors make a distinction between three components of the competency trap: ‘familiarity trap’, ‘maturity trap’ and ‘propinquity trap’. 7. This rate increases with the technological intensity of the sector studied. Thus, the cooperation rate (ranging from the creation of joint ventures to long-term client–supplier relationships) in the metal industry is 11 per cent, against the all-industry average of 15 per cent.

REFERENCES Acs, Z.J. and A. Varga (2005), ‘Entrepreneurship, agglomeration and technological change’, Small Business Economics, 24(3), 323–34. Ahuja, G. and C.M. Lampert (2001), ‘Entrepreneurship in the large corporation:

Persistence of old habits and renewal of knowledge base in French SMEs 271 a longitudinal study of how established firms create breakthrough inventions’, Strategic Management Journal, 22(7), 521–44. Aldrich, H.E. (1999), Organizations Evolving, London, UK: Sage. Bandura, A. (1986), Social Foundations of Thought and Action: A Social Cognitive Theory, Englewood Cliﬀs, NJ, USA: Prentice-Hall. Bargh, J. (1997), ‘The automaticity of every day life’, in John A. Bargh and Robert S. Wyer Jr (eds), The Automaticity of Everyday Life, Mahwah, NJ, USA, Lawrence Erlbaum Associates, pp. 1–61. Becker, M.C. and N. Lazaric (2003), ‘The influence of knowledge in the replication of routines’, Economie Appliquée, 16(3) 65–95. Becker, M.C., T. Knudsen and J.G. March (2006), ‘Schumpeter, Winter, and the sources of novelty’, Industrial and Corporate Change, 15(2), 353–71. Becker, M.C., N. Lazaric, R.R. Nelson and S.G Winter (2005), ‘Toward an operationalisation of the routines concept’, Industrial and Corporate Change, 14(5), 775–91. Beise-Zee, R. and C. Rammer (2006), ‘Local user–producer interaction in innovation and export performance of firms’, Small Business Economics, 27(2/3), 207–22. Boulding, K. (1956), The Image, Michigan, USA: University of Michigan Press. Brette, O. (2003), ‘Thorstein Veblen’s theory of institutional change: beyond technological determinism’, The European Journal of History of Economic Thought, 10(3), 455–77. Cattani, G. (2005), ‘Preadaptation, firm heterogeneity, and technological performance: a study on the evolution of fiber optics, 1970–1995’, Organization Science, 16, 563–80. Cohen, M.D. and P. Bacdayan (1994), ‘Organisational routines are stored as procedural memory: evidence from a laboratory study’, Organization Science, 5(4), 554–68. Cohen, M.D., R. Burkhart, G. Dosi, M. Egidi, L. Marengo, M. Warglien and S.G. Winter (1996), ‘Routines and other recurring action patterns of organizations: contemporary research issues’, Industrial and Corporate Change, 5(3), 653–97. Cohen, W.M. and D.A. Levinthal (1990), ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35, 128–52. Commons, J.R. (1934), Institutional Economics: Its Place in the Political Economy, New Brunswick, USA: Transaction Publishers. Commons, J.R. (1950), The Economics of Collective Action, New York, USA: Macmillan Costello, N. (1996), ‘Learning and routines in high-tech SMEs: analyzing rich case study material’, Journal of Economic Issues, 30(2), 591–7. Davenport, S. (2005), ‘Exploring the role of proximity in SME knowledgeacquisition’, Research Policy, June, 43(5), 683–701. Demers, C. (2007), Organizational Change Theories: A Synthesis, London, UK: Sage. Dopfer, K. (2007), ‘The evolutionary foundations of behavioural economics: Leibenstein’s legacy’, in R. Frantz (ed.) Renaissance in Behavioural Economics: Essays in Honor of Harvey Leibenstein, London, UK: Routledge, pp. 59–91. Dutraive, V. (2009), ‘The pragmatist view of knowledge and beliefs in institutional economics: the significance of habits of thought, transactions and institutions in the conception of economic behaviors’, in R. Arena, A. Festre and N. Lazaric (eds), Handbook of Economics and Knowledge, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, forthcoming.

272

Organizational routines and organizational change and innovation

Endres, A.M. and C.R. Woods (2006), ‘Modern theories of entrepreneurial behavior: a comparison and appraisal’, Small Business Economics, 26(2), 189–202. Feldman, M.S. (2000), ‘Organizational routines as a source of continuous change’, Organization Science, 11(6), 611–29. Feldman, M.S. (2004), ‘Resources in emerging structures and processes of change’, Organization Science, 15(3), May–June, 295–309. Feldman, M.S. and B.T. Pentland (2003), ‘Reconceptualizing organizational routines as a source of flexibility and change’, Administrative Science Quarterly, 48, 94–118. Fukugawa, N. (2006), ‘Determining factors in innovation of small firm networks: a case of cross industry groups in Japan’, Small Business Economics, 27(2/3), 181–93. Giddens, A. (1984), The Constitution of Society: Outline of the Theory of Structure, Berkeley, CA, USA: University of California Press. Greve, H.R. (1998), ‘Performance, aspirations, and risky organizational change’, Administrative Science Quarterly, 43, 58–86. Greve, H.R. (1999), ‘The eﬀect of change on performance: inertia and regression toward the mean’, Administrative Science Quarterly, 44(September), 590–614. Greve, H.R. (2007), ‘Exploration and exploitation in product innovation’, Industrial and Corporate Change, 16(5), 945–75. Hannan, M. and J. Freeman (1984), ‘Structural inertia and organizational change’, American Sociological Review, 49, 149–64. Hodgson, G.M. (1988), Economics and Institutions: A Manifesto for a Modern Institutional Economics, Cambridge, UK, Polity Press. Hodgson, G.M. (1993), Economics and Evolution: Bringing Life back into Economics, Cambridge, UK: Polity Press. Hodgson, G.M. (2007), ‘Institutions and individuals: interaction and evolution’, Organization Studies, 28(1), 95–116. Howard-Grenville, J.A. (2005), ‘The persistence of flexible organizational routines; the role of agency and organizational context’, Organization Science, 6, 618–36. Huet, F. (2004), ‘Apprentissage collectif et dynamique coopérative: une étude empirique des PME françaises’, Doctoral PhD, University of Technology of Compiègne (France). Huet, F. and N. Lazaric (2004), ‘Apprentissage coopératif et complémentarité des mécanismes de coordination: une étude empirique’, Economies et Sociétés, 38(12), 2073–105. Huet, F. and N. Lazaric (2008), ‘Capacité d’absorption et d’interaction: une étude de la coopération dans les PME françaises’, Revue d’économie Industrielle, 121, 65–84. Kahneman, D. (2003), ‘Maps of bounded rationality: a perspective on intuitive judgment and choice’, in T. Frangsmyr (ed.), Les Prix Nobel 2002 [Nobel Prizes 2002], Stockholm, Sweden: Almquist & Wiksell International. Kirzner, I.M. (1979), Perception, Opportunity and Profit, Chicago, IL, USA: University of Chicago Press. Langer, E.J. and M. Moldoveanu (2000), ‘The construct of mindfulness’, Journal of Social Issues, 56(1), 1–9. Lazaric, N. (2000), ‘The role of routines, rules and habits in collective learning: some epistemological and ontological considerations’, European Journal of Economic and Social Systems, 14(2), 157–72. Lazaric, N. (2008), ‘Routines and routinization: an exploration of some micro-

Persistence of old habits and renewal of knowledge base in French SMEs 273 cognitive foundations’, in M. Becker (ed.), Handbook of Organizational Routines, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 205–27. Lazaric, N. and B. Denis (2005), ‘Routinisation and memorisation of tasks inside a workshop: an illustration through a case study’, Industrial and Corporate Change, 14(5), 873–96. Lazaric, N., C. Longhi and C. Thomas (2008), ‘Gatekeepers of knowledge versus platforms of knowledge: an illustration with the case of a high tech cluster’, Regional Studies, 3(2), 45–63. Levinthal, D.A. and J.G. March (1993), ‘The myopia of learning’, Strategic Management Journal, 14, 95–112. Liao, J., H. Welsch and M. Stoica (2003), ‘Organizational absorptive capacity and responsiveness: an empirical investigation of growth-oriented SMEs’, Entrepreneurship Theory and Practice, 28(1), 63–84. Lorenz, E. (2000), ‘Organisational routines in the light of “old” evolutionary economics: bringing politics back into the study of organisational learning’, European Journal of Economic and Social Systems, 14(2), 191–207. Maskell, P. and A. Malmberg (2007), ‘Myopia, knowledge development and cluster evolution’, Journal of Economic Geography, 7(5), 603–18. Mueller, P. (2007), ‘Exploiting entrepreneurial opportunities: the impact of entrepreneurship on growth’, Small Business Economics, 28(4), 355–62. Nelson, R.R. (1994), ‘Routines’, in G. Hodgson, W. Samuels and M. Tool (eds), The Elgar Companion to Institutional and Evolutionary Economics, vol. 2, Aldershot, UK and Brookfield, USA: Edward Elgar, pp. 249–53. Nelson, R.R. (1995), ‘Recent evolutionary theorizing about economic change’, Journal of Evolutionary Economics, 12, 17–28. Nelson, R.R. and B.N. Sampat (2001), ‘Making sense of institutions as a factor shaping economic performance’, Journal of Economic Behavior & Organization, 44, 31–54. Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA, USA: Harvard University Press. Nooteboom, B. (1994), ‘Firm size eﬀects on transaction costs’, Small Business Economics, 5(4), 283–95. Nooteboom, B. (2000), ‘Learning by interaction, absorptive capacity, cognitive distance and governance’, Journal of Management and Governance, 4, 69–92. Pacitto, J.C. and F. Tordjman (1999), ‘L’innovation technologique dans la très petite entreprise industrielle française: ce que disent les statistiques’, Revue Internationale PME, 12(3), 59–90. Patalano, R. (2009), ‘Imagination and perception as gateways to knowledge. The unexplored aﬃnity between Boulding and Hayek’, in R. Arena, A. Festre and N. Lazaric (eds), Handbook of Economics and Knowledge, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, forthcoming. Pentland, B.T. and M.S. Feldman (2005), ‘Organizational routines as a unit of analysis’, Industrial and Corporate Change, 14(5), 793–815. Porac, J. and H. Thomas (1990), ‘Taxonomic mental models in competitor definition’, Academy of Management Review, 15, 224–40. Salvato, C. (2003), ‘The role of micro-strategies in the engineering of firm evolution’, Journal of Management Studies, 40, 83–108. Santarelli, A. and A. Sterlacchini (1990), ‘Innovation, formal vs informal R&D and firm size: some evidence from Italian manufacturing firms’, Small Business Economics, 2, 223–8.

274

Organizational routines and organizational change and innovation

Schumpeter, J.A. (1934), The Theory of Economic Development. An Inquiry into Profits, Capital, Credit, Interest, and the Business Cycle, Cambridge, MA, USA: Harvard University Press. Schumpeter, J.A. (1942), Capitalism, Socialism and Democracy, New York, USA: Harper and Brothers. Tether, B.S. (2002), ‘Who co-operates for innovation, and why. An empirical analysis’, Research Policy, 31, 947–67. Veblen, T. (1898), Why is Economics Not an Evolutionary Science?, New York, USA: The Viking Press. Veblen, T. (1904), The Theory of Business Enterprise, New York, USA: Charles Scribners Sons. Veblen, T. (1914), The Instinct of Workmanship and the State of the Industrial Arts, reprint with a new Introduction by Murray G. Murphey, New Brunswick, NJ, USA and London, UK: Transaction Publishers [1990]. Veblen, T. (1919), The Place of Science in Modern Civilisation and Other Essays, New York, USA: Huebsch. Vromen, J.J. (2006), ‘Routines, genes and program-based behavior’, Journal of Evolutionary Economics, 16(5), 543–60. Weick, K.E. (1995), Sensemaking in Organization, Thousand Oaks, CA, USA: Sage Publications. Weick, K.E. (2000), Making Sense of the Organization, Oxford, UK: Blackwell Business. Weick, K.E. and K.M. Sutcliﬀe (2006), ‘Mindfulness and the quality of organizational attention’, Organization Science, 17(4), 514–24. Winter, S.G. (1975), ‘Optimization and evolution in the theory of the firm’, in R.H. Day and T. Groves (eds), Adaptive Economic Models, New York, USA: Academic Press pp. 73–118. Witt, U. (2000), ‘Changing cognitive frames – changing organizational forms: an entrepreneurial theory of organizational development’, Industrial and Corporate Change, 9(4), 733–55. Zouhour, K. (2003), ‘Compétences pour innover et coopérations technologiques – Une analyse multivariée de l’industrie française’, Revue d économie Industrielle, 102, 29–54.

Persistence of old habits and renewal of knowledge base in French SMEs 275

APPENDIX 11.1 Table A11.4

THE QUALITATIVE INTERVIEWS CONDUCTED

Interviews conducted in the framework of this study

Firm code

Number of Region employees

AA1

17

AA2

10

AA3

60

AA4

100

AA5

48

AA6

113

AA7

30

CE1

40

CE2

70

SS1

30

TM1

20

TM2

21

TM3

15

Burgundy

Main activity

Main Size of the partner or partner partnership firm(s)

Food Supplier processingconfectionery Provence- Casing Horizontal Alpes Cote processing d’Azur Region Brittany Smoked Supplier products The Loire Pre-cooked Supplier Region meals Greater TorrefacClient Paris tion Region Brittany Pre-cooked Client meals RhôneBakeryClient Alpes food Region processing Greater Printed Supplier Paris circuits Region Aquitaine Electronic Commerservices cial joint venture Greater I.T. Client Paris developRegion ment Picardy Metal Commercutting and cial pressing association Picardy Plastic Client casing Picardy Galvanisa- Economic tion Interest Group

Location of the partner(s)

Bigger

Same region

Bigger

Morocco

Similar size Similar size Bigger

Same region

Same region

Bigger

France

Similar size

Same region

Similar size

Same region

Similar size

France

Similar size

Same region

Similar size

Same region

Bigger

Same region

Similar size

Same region

France

276

Organizational routines and organizational change and innovation

Table A11.4

(continued)

Firm code

Number of Region employees

Main activity

Main Size of the partner or partner partnership firm(s)

Location of the partner(s)

TM4

21

Picardy

Same region

10

Alsace

Economic Interest Group Client

Similar size

TM5

Bigger

Same region

TM6

10

TM6

10

RhôneAlpes Region Picardy

TM7

75

PolishingNickel plating Sheet metal work Shank manufacturing Engineering consulting Aeronautics repairs

TM8

50

TM9

10

TM10

35

Precision instrument industry Valves and fittings industry Aeronautical mechanics

RhôneAlpes Region Greater Paris Region Picardy

MidiPyrénées Region

Horizontal Bigger

Bulgaria

SPL

Bigger

Same region

Client

Bigger

Same region

Client / Bigger Horizontal

Same region

Client

Same region

Bigger

Horizontal Similar (Economic size Interest Group)

Same region

Index Abbott, A. 57, 64, 69, 72, 74, 76, 83, 84, 93 Abelson, R.P. 162, 163 absorptive capacity 257–8, 260–61, 269 Acs, Z.J. 248 action dispositions 143–6, 153–6 see also dispositions action network 58–60 action patterns 49, 50–51, 132, 154–5, 161, 162 Adler, P.S. 153, 218, 224, 241 adverse news routine 170–73 Agarwal, R. 104 agencements 4, 195, 211–12, 214–15, 217–18 agencies 190, 209–11, 212, 214–15 and calculation 193–4, 207 see also artefacts and agencies, influence on routines’ dynamics Ahuja, G. 238, 263 Aiken, M. 224 Akrich, M. 191 Aldenderfer, M.S. 85 Aldrich, H.E. 36, 268 Alessi product development routines, event-sequence analysis 77–84 analysis and findings 84–90 cluster 1 recipe-book projects 85–6 cluster 2 in-house mutated projects 86–7 cluster 3 externally mutated projects 87–8 cluster 4 recombinant projects 88–9 cluster 5 unconventional projects 89 heterogeneity of routine patterns 85–9 routines evolution 89–90 cluster analysis 83–4 data analysis 79–84 data sources 78–9

empirical setting 77–8 optimal matching analysis (OMA) 72–7, 83, 90–91 Allison, G.T. 154 Ambady, N. 143 ambiguity between rules and routines 159–60, 177, 179–80 Rabobank, case study 166–7 adverse news routine 170–73 control routine, absence of 173–7 methodology 167, 169 rules, routines and ambiguity 169– 77, 178–9 theoretical background 160–66 inertia through ambiguity 166 routines 161–2 rules 160–61 scripts 162–5 consciously invoked 164 as individual resources 164–5 tacitly invoked 163–4 Anderson, J.R. 154 Anderson, N. 224 Andrews, J. 224 Apple Computers 223 Aristotle 31 artefacts and agencies, influence on routines’ dynamics 4, 185–7 conclusions 213–16 discussion and framework performativity 205–9 performativity struggles and agency 210–12 prescription, description, performativity and artefacts 209–10 production development routine (freeze process) 196–205 actual process 198–204 formal process 197–8 mutual adaptation 204–5 software introduction 195–6

277

278

Index

theoretical foundation 188–95 description vs. prescription 189–92 performation 192–5 routines, artefacts and agency 188–92 rule-following 191–2 theories and reality relationship 193–5 Augier, A. 206 Aunger, R. 36 Axelrod, R.M. 146 Bacdayan, P. 33–4, 35, 37, 144, 161, 251 Bakeman, R.J. 79, 83 Ball, A. 147 Ball, B. 147 Balogun, J. 169 Bandura, A. 250 Banks, M. 110, 118–20 Bargh, J. 251 Barley, S.R. 50, 160, 162, 190 Barnes, B. 190 Barnes, W. 166 Barnesian performativity 194, 208, 210 Basu, A. 47 Bawa, P. 114, 117 Bechky, B.A. 217 Becker, J. 47 Becker, M.C. 1, 3, 4, 5, 36, 40, 47, 51, 69, 70, 71, 72, 103, 112, 161, 162, 164, 185, 217, 218, 224, 234, 243, 248, 252, 253, 268 behaviour, diffusion of 36–7 behaviour patterns 162, 172–3 behaviourist psychology 27 Beise-Zee, R. 256 Bender-deMoll, S. 60 Benghozi, P.-J. 224, 243 Berg, M. 189, 217 Berger, P. 162 Beyer, J.M. 146 Bhaskar, R. 40 Bill of Materials (BoM) freeze process 187, 195–206, 212 actual process 198–204 formal process 197–8 mutual adaptation 204–5

see also artefacts and agencies, influence on routines’ dynamics Birnholtz, J. 146 black box approach 69, 114, 125, 185 Blackler, F. 153 Blair-Loy, M. 83, 84, 85 Blashfield, R.K. 85 Blau, P. 209 Blitz, D. 35, 40 Bloor, D. 190, 217 Blossfeld, H.-P. 73 blueprints 114, 116, 126 Bohn, R.E. 238 BoM (Bill of Materials) freeze process 187, 195–206, 212 actual process 198–204 formal process 197–8 mutual adaptation 204–5 see also artefacts and agencies, influence on routines’ dynamics Borys, B. 224 Boschma, R.A. 104, 105 Boulding, K. 250, 251 Bourdieu, P. 189, 192, 217 Bowker, G.C. 206 Boyd, R. 28, 29 Brette, O. 254 Bridges, W. 156 Brown, J.S. 217, 218 Bruce, R.A. 112, 224 Bucciarelli, L.L. 217 Buenstorf, G. 113 Bunge, M.A. 40 Burns, J.E. 159, 160, 161, 162, 180 Burns, T. 224 Burton, M.D. 115 Busco, C. 172 Business of Systems Integration (Prencipe, Davies, Hobday) 19 Butts, C.T. 65 calculative agency 193–4, 207 Callon, M. 186, 193, 195, 207, 211–12 Camic, C. 40 Camp Poplar Grove regeneration 132–3 observation 133–42 data gathering and analysis 134, 136 observations summary 136–42

Index regenerative processes 147–53 communication 150–52 demonstration 147–8 generic skills application 152–3 guidance cascade 148–9 sketch map 135 Campbell, D.T. 29 Cantner, U. 104 capabilities 11–12, 19, 31, 33, 35, 116, 126 capacities 31, 33, 35, 37 car manufacturing innovation case study 223–5, 239–41 analytical perspective and method 228–30 empirical findings, virtual simulation tools 230–39 innovation outcomes 232–9 opportunities provided 233–4 procedures’ leverage of opportunities 234–7 setting up 230–31 stability, effects of 237–9 use of 231–2 literature 226–7 new product development 227–8 organizational routines 225–6 Cardinal, L.B. 224 cascade of guidance 137, 148–9 Cassell, P. 190 Cattani, G. 253 Chandler, A.D. Jr. 24 Chapon, F. 108 Chi, M.T.H. 48 Chia, R. 153, 155 Clark, K.B. 224, 228, 234 cluster analysis 73, 74, 77, 83–4 see also event-sequence analysis of routines codification 206–7 cognitive automatisms 251 cognitive distance 262–3 cognitive memory systems 34 cognitive regularities 70, 249–51, 265–6 cognitive scripts 162–3 see also scripts cognitivist view 189 Cohen, J. 83 Cohen, M.D. 1, 29, 33–4, 35, 37, 70,

279

144, 146, 161, 185, 187, 189, 225, 251, 267 Cohen, W.M. 257 Cole, R. 23 collective action dispositions 144 see also dispositions collective actions patterns 132, 161, 162, 251–2 collective communication bursts 150– 51 collective sensemaking process 171–2 Collins, H.M. 217 combinatorics of routines 3, 6, 62–3, 253, 268 Commons, J.R. 5, 248, 251–2 communication in regenerative process 150 collective bursts 150–51 individual bursts 151–2 computer program analogy 30 consciously invoked scripts 163, 164, 165, 173, 177, 179 contagious behaviour 36–7 convergence between formal procedure and performance 208–9, 211, 213 Cooper, R.G. 223, 227, 228 cooperation 253, 261–2, 263–4, 265, 267, 269 Coriat, B. 218 Cornford, J. 192 Costello, N. 248 counter-performativity 194–5, 209, 210, 214 creative industries 104, 105 creativity and routinization 105, 111– 13 see also inheritance of fashion industry routines Creed, W.E.D. 160 Cuban missile crisis, soldiers’ established practice 154 Cyert, R.M. 68, 70, 79, 185, 224, 228 D’Adderio, L. 185, 189, 191, 192, 206, 211, 216, 217 Daft, R.L. 190 Dahl, M.S. 104 Dale, A. 73 Damanpour, F. 224, 241 Darwin, C.R. 35

280

Index

Davenport, S. 249 Davies, A. 19 Davila, T. 224 Davis, R.B. 73 De Jong, J.P.J. 224 De Marly, D. 106, 107, 114, 117 De Rethy, E. 117 declarative memory 154 Delbecq, A.L. 224 Demers, C. 268 demonstration in regeneration process 147–8 Denis, B. 36, 68, 185, 270 Devereux, J. 64 Dewey, J. 27, 155 diachronic comparison 50, 51 diffusion of behaviour 36–7 DiMaggio, P.J. 36 Dimcock, H.S. 147 disentanglement 193, 206 dispositions 2, 30, 33, 39, 142, 162 action dispositions 143–6, 153–6 distributed agencies 210–11, 214–15 divergence of performance from formal procedure 209, 211, 213 DNA sequencing 73–4 Dopfer, K. 251 Dosi, G. 29, 68, 79, 166, 218, 243 Doucet, Jacques 108–9 Drought, R.A. 147 Druckman, D. 146 Duguid, P. 217, 218 Dutraive, V. 252 dynamic networks 60–61 economic markets, performativity 193–5 effective performativity 194, 208, 210 Egidi, M. 166 Eisenhardt, K.M. 68, 78, 167, 187, 229 Eisner, M.D. 133 emergent properties 30 Emerson, R.M. 134 emotions 155 empirical knowledge 154 Employee Value Plan 175, 176 Endres, A.M. 248 entrepreneurs 250–51, 252–3, 265, 266–7 established practices 154 Evans, M.G. 40

event history models 73 event logs see workflow event logs event-sequence analysis of routines 69–72, 90–92 Alessi product development routines 77–84 analysis and findings 84–90 cluster 1 recipe-book projects 85–6 cluster 2 in-house mutated projects 86–7 cluster 3 externally mutated projects 87–8 cluster 4 recombinant projects 88–9 cluster 5 unconventional projects 89 heterogeneity of routine patterns 85–9 routines evolution 89–90 data analysis 79–84 data sources 78–9 empirical setting 77–8 cluster analysis 83–4 limitations 91 optimal matching analysis (OMA) 72–7, 83, 90–91 algorithm 74–7 in DNA sequencing 73–4 evolution of routines 13–16 physical technologies, 17–18 hybrid corn 17–18, 20–21 polio vaccines 17–18, 21–2 social technologies, 17, 18 M form 17, 18–19, 24–5 quality circles 17, 18–19, 22–3 understanding, strength of 16, 17–20 see also event-sequence analysis of routines experimentation 16, 228–9, 242 see also innovation, role of procedures and stable behaviour patterns exploitation 224, 242, 253 exploration 224, 238, 240, 241, 242, 253, 264 Fanelli, A. 153, 154 fashion design industry, genealogy and routine inheritance 3

Index analysis conclusions 124–6 data collection 117–18 hypotheses 115–17 methodology 118–20 results 120–24 spin-off mechanism of routine replication inheritance 113–17 routine and creativity dualism 111–13 genealogy 109–11 history 106–11 haute couture 106–7 ready-to-wear 107, 109–10 spin-offs 107–11 Faust, K. 60 Feldman, D.C. 147 Feldman, M.S. 6, 69, 70, 71, 84, 89, 91, 92, 112, 114, 125, 153, 154, 161, 165, 166, 185, 186, 188, 194, 215, 225, 241, 248, 251 financial markets, performativity 193–5 Fincham, J.M. 154 Fiol, C.M. 146 Fleming, L. 238 flowcharts 50 Ford mass production 12 recall schema 164 Ford, C.M. 112, 224 formal rules 161, 165, 175–7, 188, 210 framing 186–7, 193–4, 195, 206 framing, overflowing and reframing cycles 186–7, 209, 211, 213 framing view 189, 195 Freeman, J. 36, 37, 38, 250 freeze process see artefacts and agencies, influence on routines’ dynamics; Bill of Materials (BoM) freeze process French SMEs, novel routines and organizational conservatism 5, 248–9, 269–70 analytical framework 249–54 automaticity and mindful attitude 251–4 images 249–51 discussion, change and stability sources 267–8

281

observation of interactions in SMEs 254–64 data presentation and methodology 254–9 qualitative investigation 265–7 cognitive regularities and new frames 265–6 entrepreneurs and public institutions affecting change 266–7 quantitative results and analyses 259–64 cognitive distance and knowledge absorption 262–3 cooperation and knowledge base renewal 263–4 econometric results 259–60 initial conditions, path dependency and absorptive capacity 260–61 interaction intention and information circulation 261–2 Friesen, P.H. 166 Fujimoto, T. 237 Fukugawa, N. 258 Galison, P. 211 Garfinkel, H. 190 Gasser, L. 202, 217 Gasteiger, R.M. 224 Gavetti, G. 166 gene analogy 29–33 genealogy and routine inheritance in the fashion industry 3 analysis conclusions 124–6 data collection 117–18 hypotheses 115–17 methodology 118–20 results 120–24 spin-off mechanism of routine replication inheritance 113–17 routine and creativity dualism 111–13 genealogy 109–11 history 106–11 haute couture 106–7

282

Index

ready-to-wear 107, 109–10 spin-offs 107–11 generative systems 48–9, 69 generic performativity 194, 208, 210 Gibson, H.W. 147 Giddens, A. 160, 190, 252 Gilbert, C.G. 166 Gilfallan, D.P. 146 Gilson, L.L. 224 Gioia, D.A. 112, 162, 163, 164, 224 Glaser, B. 187 Godfrey-Smith, P. 36 Goodrich, L. 147 Gort, M. 104 Gottman, M. 79, 83 Greve, H.R. 241, 251, 253 Gribskov, M. 64 Griffin, A. 224, 228 Grint, K. 191 guidance cascade 137, 148–9 Gupta, A.K. 224 habits 26–9, 34, 39 replication of 28–9, 37, 38 Haerem, T. 48 Hamilton, A.E. 147 Hannan, M.T. 36, 37, 38, 250 Hargadon, A. 153, 154, 223, 224 Harré, R. 40 Harris, S.G. 163–4 Hatherly, D. 190, 191, 192 Hauser, J.R. 224, 228 haute couture 106–7 see also high fashion design industry, routine inheritance Helfat, C.E. 104, 114, 115 Hennion, A. 194 Henrich, J. 29 high fashion design industry, routine inheritance 3 analysis conclusions 124–6 data collection 117–18 hypotheses 115–17 methodology 118–20 results 120–24 spin-off mechanism of routine replication inheritance 113–17

routine and creativity dualism 111–13 genealogy 109–11 history 106–11 haute couture 106–7 ready-to-wear 107, 109–10 spin-offs 107–11 Hobday, M. 19 Hodgson, G.M. 27, 28, 33, 35, 38, 39, 40, 103, 161, 162, 225, 226, 248, 250, 252 Holm, P. 194 Howard-Grenville, J.A. 50, 254 Hrycak, A. 64, 69, 74 Huet, F. 254, 261, 262 Hull, D.L. 36, 38 Humphreys, P. 40 Hutchins, E. 191, 216, 217 hybrid corn, case study 17–18, 20–21 Iansiti, M. 206 identification of routines 48, 51 IDEO 223, 224, 240 images 250–51, 253 imitation 28–9 improvisation 87, 88, 89–90, 112, 152–3, 161 incentive 28, 175–6, 179 individual communication bursts 151–2 industry life cycles 104–5 inertia between formal rules and routines 159–60, 177, 179–80 Rabobank, case study 166–7 adverse news routine 170–73 control routine, absence of 173–7 methodology 167, 169 rules, routines and ambiguity 169–77, 178–9 theoretical background 160–66 inertia through ambiguity 166 routines 161–2 rules 160–61 scripts 162–5 consciously invoked 164 as individual resources 164–5 tacitly invoked 163–4 information systems 191–2 information transfer 33–5, 36, 37–8, 147–53

Index communication 150–52 demonstration 147–8 guidance cascade 148–9 inheritance of fashion industry routines 3 analysis conclusions 124–6 data collection 117–18 hypotheses 115–17 methodology 118–20 results 120–24 spin-off mechanism of routine replication inheritance 113–17 routine and creativity dualism 111–13 genealogy of the industry 109–11 history of the industry 106–11 haute couture 106–7 ready-to-wear 107, 109–10 spin-offs 107–11 inheritance of organizational routines 103–5 see also inheritance of fashion industry routines innovation evolution, case studies hybrid corn 17–18, 20–21 M form 17, 18–19, 24–5 polio vaccines 17–18, 21–2 quality circles 17, 18–19, 22–3 innovation, role of procedures and stable behaviour patterns 223–5, 239–41 analytical perspective and method 228–30 empirical findings, virtual simulation tools 230–39 innovation outcomes 232–9 opportunities provided 233–4 procedures’ leverage of opportunities 234–7 setting up 230–31 stability, effects of 237–9 use of 231–2 literature 226–7 new product development 227–8 organizational routines 225–6 instincts 28, 29 institutional change 159, 160 institutionalization of cooperation 267

283

institutionalized mind 251–4 intentionality 28, 89–90 interaction between rules and routines 159–60, 177, 179–80 Rabobank, case study 166–7 adverse news routine 170–73 control routine, absence of 173–7 methodology 167, 169 rules, routines and ambiguity 169–77, 178–9 theoretical background 160–66 inertia through ambiguity 166 routines 161–2 rules 160–61 scripts 162–5 consciously invoked 164 as individual resources 164–5 tacitly invoked 163–4 interaction capacity 256, 257, 262–3 interaction patterns 69, 162–3, 164–5 interactors 38–9 invoice processing 51–2, 62 see also workflow data, use in routine structure analysis Isabella, L.A. 79, 82 iterative minimization procedure 74–7 Jablin, F.M. 146, 147, 152 Jacobs, J. 148 James, W. 27, 29 Jelinek, M. 224 Johnson, G. 169 Johnson, J.D. 224 Kahn, K.B. 227, 228 Kahneman, D. 253 Kandel, E.R. 144, 154 the Kappa way 155 Kaspar, F. 65 Kasparow, Garri 207 Katila, R. 238 Kemp, R. 224 Kilpinen, E. 40 Kim, J. 40 Kirzner, I.M. 253 Kleinschmidt, E.J. 223, 227 Klepper, S. 103, 104, 105, 109, 114, 115, 116, 117, 122, 125 knowing 154

284

Index

knowledge transfer 33–5, 36, 37–8, 147–53 communication 150–52 demonstration 147–8 guidance cascade 148–9 Knudsen, T. 29, 31, 38, 39, 40, 103, 161, 162, 225, 226 Kogut, B. 68 Kolgomorov, A.N. 65 Kontopoulos, K.M. 35, 40 Kripke, S.A. 178 Kruskal, J.B. 57, 64, 74, 76, 83 Kumar, A. 47 Labianca, G. 169 Lampert, C.M. 263 Langer, E.J. 254 latent knowledge 154 Latour, B. 191, 217, 218 Lave, J. 147, 214 Lazaric, N. 36, 40, 68, 70, 161, 185, 248, 251, 254, 261, 268, 270 Leiken, J. 147 Lempel, A. 65 Lempel-Ziv complexity 57, 65 Levenshtein distance 57, 64, 83 Levinthal, D.A. 238, 241, 253, 257 Levitt, B. 33, 36, 68 lexical variety measures 54 Liao, J. 249, 269 Lieberman, M.B. 104, 114, 115 localized learning 253 longitudinal data analysis 73 Lorenz, E. 248 Louis, M.R. 162, 164 Lowe, A. 177 Luckmann, T. 162 Lynch, M. 190, 191 M form 17, 18–19, 24–5 MacIndoe, H. 69, 72, 76 MacKenzie, D. 186, 194–5, 208, 210, 215 Mackey, A. 146 Madden, E.H. 40 Malmberg, A. 253 Mambrey, P. 217 March, J.G. 33, 35, 36, 68, 70, 79, 185, 206, 224, 228, 236, 253

Marengo, L. 206 Margolis, H. 40 market models 194 Martin, J.A. 68, 146 Martin, R. 117 Martinez, M. 36 Maskell, P. 253 Mathieu, J.E. 224 McDougall, W. 27 McFarland, D. 60 McIntosh, A. 177 mechanistic view 189, 209 Mellor, S. 224 Metaphysics (Aristotle) 31 methodology 5–6 Miller, D. 166 Miller, V.D. 146, 147, 152 mindful reflexivity 254 Mintzberg, H. 79, 83 Moldoveanu, M. 254 Moody, J. 60 Morison, E.E. 35 Murphy, J.B. 40 Narduzzo, A. 166 Needleman-Wunsch algorithm 83 Nelson, R.R. 12, 29, 30, 33, 37, 51, 68, 70, 74, 79, 84, 90, 103, 111–12, 114, 166, 215, 224, 225, 226–7, 228, 248 network graphs 58–60 network models 51 new product development routines, event-sequence analysis 69–72, 90–92 Alessi product development routines 77–84 analysis and findings 84–90 cluster 1 recipe-book projects 85–6 cluster 2 in-house mutated projects 86–7 cluster 3 externally mutated projects 87–8 cluster 4 recombinant projects 88–9 cluster 5 unconventional projects 89 heterogeneity of routine patterns 85–9

Index routines evolution 89–90 data analysis 79–84 data sources 78–9 empirical setting 77–8 cluster analysis 83–4 optimal matching analysis (OMA) 72–7, 83, 90–91 algorithm 74–7 in DNA sequencing 73–4 new product development, virtual simulation technology 223–5, 239–41 analytical perspective and method 228–30 empirical findings 230–39 innovation outcomes 232–9 opportunities provided 233–4 procedures’ leverage of opportunities 234–7 setting up 230–31 stability, effects of 237–9 use of 231–2 literature 226–7 organizational routines 227–8 Newell, A. 69 newness, fear of 250, 253 NK models 238 non-performativity 209, 210 Nooteboom, B. 165, 249, 250 novel routine creation and organizational conservatism 5, 248–9, 269–70 analytical framework 249–54 automaticity and mindful attitude 251–4 images 249–51 discussion, change and stability sources 267–8 observation of interactions in SMEs 254–64 data presentation and methodology 254–9 qualitative investigation 265–7 cognitive regularities and new frames 265–6 entrepreneurs and public institutions affecting change 266–7 quantitative results and analyses 259–64

285 cognitive distance and knowledge absorption 262–3 cooperation and knowledge base renewal 263–4 econometric results 259–60 initial conditions, path dependency and absorptive capacity 260–61 interaction intention and information circulation 261–2

obsolete knowledge in routines 35 Ocasio, W. 236 O’Hara Callan, G. 107, 108, 110, 111, 117, 118, 124 Ohly, S. 105, 112, 224 optimal matching analysis (OMA) 69, 72–7, 83, 90–91 algorithms 74–7, 83 in DNA sequencing 73–4 optimal string matching 57, 64 OPTIMIZE© 83 O’Reilly, C. 224 organizational blueprints 114, 126 organizational capabilities 11–12, 19, 31, 33, 35, 116, 126 organizational character 142–3 organizational character regeneration 131–3 action dispositions 143–4 and mutual adaptation 145–6 persistence of 153–6 Camp Poplar Grove 132–3 observation 133–42 data gathering and analysis 134, 136 observations summary 136–42 regenerative processes 147–53 communication 150–52 demonstration 147–8 generic skills application 152–3 guidance cascade 148–9 character in relation to other perspectives 146–7 organizational character 142–3 procedural memory 144

286

Index

organizational conservatism and novel routine creation 5, 248–9, 269–70 analytical framework 249–54 automaticity and mindful attitude 251–4 images 249–51 discussion, change and stability sources 267–8 observation of interactions in SMEs 254–64 data presentation and methodology 254–9 qualitative investigation 265–7 cognitive regularities and new frames 265–6 entrepreneurs and public institutions affecting change 266–7 quantitative results and analyses 259–64 cognitive distance and knowledge absorption 262–3 cooperation and knowledge base renewal 263–4 econometric results 259–60 initial conditions, path dependency and absorptive capacity 260–61 interaction intention and information circulation 261–2 organizational memory 30 organizational relationships 192, 210 Orlikowski, W. 153, 154, 155, 190, 192 Orr, J.E. 217, 218 ostensive routines 6, 154, 251 Ouellette, J.A. 40 overflowing 186, 187, 193, 194, 211, 213 overflowing view 189–91, 195 Pacitto, J.C. 261 paradox of [n]ever changing world 131–3, 154 parent–spin-off routine transfer 103–5 see also inheritance of fashion industry routines Patalano, R. 251

patterns of interaction 69, 162–3, 164–5 Pavitt, K. 224, 227, 228 Peirce, C.S. 27 Pentland, B.T. 6, 69, 70, 71, 83, 84, 91, 92, 112, 114, 125, 153, 154, 161, 165, 166, 185, 186, 188, 194, 215, 225, 226, 241, 248, 251 performative routines 154, 251 performativity 70, 193–5, 205–10, 214 categories 194–5, 208, 210 of financial markets 193–5 performativity struggles 210–11, 215 Perreau, J.L. 117 persistence of routines 35 Pettigrew, A.M. 229 Phillips, D.J. 105, 125 physical technologies 2, 12, 13 hybrid corn 17–18, 20–21 polio vaccines 17–18, 21–2 Pickering, A. 193, 213 Pierce, A.L. 224 Piguet 109, 111 Pipart, Gerard 117 Pisano, G. 162 Polanyi, M. 28, 37, 68 polio vaccines, case study 17–18, 21–2 Pollock, N. 192, 202 Poole, P.P. 162, 163 Poplar Grove see Camp Poplar Grove regeneration Popper, K.R. 40 Porac, J.F. 162, 253 potentialities 32, 33 Powell, W.W. 36 Prasad, A. 177 Prasad, P. 177 Preda, A. 191, 211, 217 Prencipe, A. 19 prescription 195, 209, 213, 214 prêt à porter 107, 109–10 Prietula, M. 206 procedural knowledge 34, 251 procedural memory 34, 35, 39, 142, 144, 145, 154 procedures and stable behaviour patterns, role in innovation 223–5, 239–41

Index analytical perspective and method 228–30 empirical findings, virtual simulation tools 230–39 innovation outcomes 232–9 opportunities provided 233–4 procedures’ leverage of opportunities 234–7 setting up 230–31 stability, effects of 237–9 use of 231–2 literature 226–7 new product development 227–8 organizational routines 225–6 process mapping 50 process models 194 Product Data Manager (PDM) software 187, 197, 200–202, 203–4, 218 product system as physical technology 19, 20 ProM 62 properties of routines 50–51 quality circles 17, 18–19, 22–3 Quine, Willard van Orman 27 Rabobank, inertia and ambiguity case study 159–60, 177, 179–80 Rabobank, case study 166–7 adverse news routine 170–73 control routine, absence of 173–7 methodology 167, 169 rules, routines and ambiguity 169–77, 178–9 theoretical background 160–66 inertia through ambiguity 166 routines 161–2 rules 160–61 scripts 162–5 consciously invoked 164 as individual resources 164–5 tacitly invoked 163–4 Rabobank West-Zeeuws Vlaanderen 171, 172 Rabobank Zevenhuizen-Moerkapelle 170–71, 172 Rafaeli, A. 165 Rammer, C. 256 rationalist view 189

287

Rau, D. 48 ready-to-wear fashion 107, 109–10 see also fashion design industry, genealogy and routine inheritance Reber, A.S. 29 reframing 186, 187, 213 regeneration of organizational character 131–3 action dispositions 143–4 and mutual adaptation 145–6 persistence of 153–6 Camp Poplar Grove 132–3 observation 133–42 data gathering and analysis 134, 136 observations summary 136–42 regenerative processes 147–53 communication 150–52 demonstration 147–8 generic skills application 152–3 guidance cascade 148–9 character in relation to other perspectives 146–7 organizational character 142–3 procedural memory 144 regularities 249–51 Reichstein, T. 104 replication of habits 28–9, 37, 38 of routines 6, 26, 36–9, 103–5 information transmission 33–5 persistence of routines 35 routines as genes 29–33 replication of routines in the fashion industry 3 analysis conclusions 124–6 data collection 117–18 hypotheses 115–17 methodology 118–20 results 120–24 spin-off mechanism of routine replication inheritance 113–17 routine and creativity dualism 111–13 genealogy of the industry 109–11 history of the industry 106–11 haute couture 106–7

288

Index

ready-to-wear 107, 109–10 spin-offs 107–11 Rerup, C. 241 resource rigidity 166 Results Oriented Management (ROM) 167, 168, 173, 174, 175, 179 Results Oriented Rewarding 175–6 Reynaud, B. 40, 161, 180, 185 Richerson, P.J. 28, 29 Riggio, J. 147 Roberts, K.H. 164 Robey, D. 64, 69, 79, 82, 83, 84, 85 Robinson, M. 217 Rogers, E.M. 36 Rohwer, G. 73 ROM see Results Oriented Management (ROM) routine combinatorics 3, 6, 62–3, 253, 268 routine inheritance see routine replication in the fashion industry routine replication 6, 26, 36–9, 103–5 information transmission 33–5 persistence of routines 35 routines as genes 29–33 routine replication in the fashion industry 3 analysis conclusions 124–6 data collection 117–18 hypotheses 115–17 methodology 118–20 results 120–24 spin-off mechanism of routine replication inheritance 113–17 routine and creativity dualism 111–13 genealogy of the industry 109–11 history of the industry 106–11 haute couture 106–7 ready-to-wear 107, 109–10 spin-offs 107–11 routine rigidity 166 routine transactions 5, 252, 253, 254, 259, 269 routine transfer between parent and spin-off 103–5 see also routine replication in the fashion industry

routines 33, 39, 70, 161–2 ‘routines as genes’ 29–33 routinization and creativity 105, 111–13 see also routine replication in the fashion industry Rueter, H.H. 71, 226 rules 160–61 rules and routines interaction, ambiguity and inertia 159–60, 177, 179–80 Rabobank, case study 166–7 adverse news routine 170–73 control routine, absence of 173–7 methodology 167, 169 rules, routines and ambiguity 169–77, 178–9 theoretical background 160–66 inertia through ambiguity 166 routines 161–2 rules 160–61 scripts 162–5 consciously invoked 164 as individual resources 164–5 tacitly invoked 163–4 Sabherwal, R. 64, 69, 79, 82, 83, 84, 85 Sahlins, M. 195 Salvato, C. 268 Sampat, B. 12, 248 Sankoff, D. 57, 64, 74, 76, 83 Santarelli, A. 261 Scapens, R.W. 159, 160, 161, 162, 180 Schacter, D.L. 34 Schank, R.C. 162, 163 Schein, E.H. 172 schemas 162, 163–4 Schön, D.A. 234 Schoonhoven, C.B. 224 Schulz, M. 225 Schumpeter, J.A. 224, 234, 238, 250, 253, 268 Schuster, H.G. 65 Scott, R. 23 Scott, S.G. 112, 224 scripts 172–3, 191 consciously invoked 163, 164 as individual resources 164–5 tacitly invoked 163–4 search processes 238–9

Index seasonal organizations 132, 147, 149 see also Camp Poplar Grove regeneration; summer camps Selznick, P. 142–3, 144 Seo, M.-G. 160 sequence analysis 71–2 Sharp, D.J. 160 Simon, H.A. 29, 35, 36, 68, 70, 185, 206, 217, 228 Simons, K.L. 105, 117 Sischy, I. 117 skills transfer 37, 147–50 Sleeper, S. 116 Sluyterman, K. 167 SMEs, novel routines and organizational conservatism 5, 248–9, 269–70 analytical framework 249–54 automaticity and mindful attitude 251–4 images 249–51 discussion, change and stability sources 267–8 observation of interactions in SMEs 254–64 data presentation and methodology 254–9 qualitative investigation 265–7 cognitive regularities and new frames 265–6 entrepreneurs and public institutions affecting change 266–7 quantitative results and analyses 259–64 cognitive distance and knowledge absorption 262–3 cooperation and knowledge base renewal 263–4 econometric results 259–60 initial conditions, path dependency and absorptive capacity 260–61 interaction intention and information circulation 261–2 Smith, D.C. 224 social technologies 2, 12, 13 M form 17, 18–19, 24–5

289

quality circles 17, 18–19, 22–3 strategic alliances 19–20 socio-technical agencements 211–12, 214–15, 217–18 software 191–2, 195–6, 207, 211–12 SoNIA 60 SOPs see standard operating procedures (SOPs) Sorenson, O. 238 Spender, J.C. 190, 217 Sperber, D. 36 spin-off companies and routine replication 103–5 see also routine replication in the fashion industry Squire, L.R. 144, 154 stability 3–4, 6, 215–16, 240, 241, 242 see also inertia between formal rules and routines; regeneration of organizational character stable behaviour patterns 4–5, 6, 225, 236–9 Stalker, G.M. 224 standard operating procedures (SOPs) 4–5, 160–61, 186–7, 188, 205–9, 225–6, 228 see also innovation, role of procedures and stable behaviour patterns standardization 206–7 Star, S.L. 206 Staw, B.M. 69 Sterelny, K. 36 Sterlacchini, A. 261 Stinchcombe, A.L. 35, 36, 104 strategic alliances 19–20 strategic transactions 252, 253, 254, 259, 262, 264, 269 Strauss, A. 187 string matching distance 64–5 structuration 160 structure analysis using workflow data 47–8 discussion, limits and possibilities 61–4 antecedents of structures 63–4 combinatorics of organizational evolution 62–3 data, use of 61–2

290

Index

sequential variety and evolution 63 structure identification 62 methodology 51–7 data, lexical variety 54–5 sequences listing 55–7 string matching distance 64–5 structure of routines 57–61 action networks 58–60 dynamic networks 60–61 network graphs and measures 58–60 sequential variety 57–8 theory 48–51 properties and patterns over time 50–51 sub-routines 51 Suchman, L. 50, 217 summer camps 3–4, 131 see also Camp Poplar Grove regeneration Sutcliffe, K.M. 254 Sutton, R.I. 69, 223, 224 synchronic comparison 50, 51 Szulanski, G. 36 tacit knowledge 35, 37, 262 tacitly invoked scripts 163–4, 165, 166, 173, 177, 179, 180 Taylor, C. 189, 190, 217 technologies, routines as 11–12 physical technologies 2, 12, 13 hybrid corn 17–18, 20–21 polio vaccines 17–18, 21–2 social technologies 2, 12, 13 M form 17, 18–19, 24–5 quality circles 17, 18–19, 22–3 strategic alliances 19–20 technology-embedded rules 191 Teece, D. 68, 103, 162 temporal durability of routines 35 Thomas, H. 253 Thomas, N. 193 Thomas, W. 27 Thomke, S.H. 237 Tolbert, S.P. 160, 162 Tomasello, M. 29, 38 Tordjman, F. 261 Toyota, lean production 12 Trice, H.M. 146

Tripsas, M. 166 Tsoukas, H. 153, 155, 189, 217 Tulving, E. 34 Turner, S.F. 227 Tushman, M.L. 224 understanding, strength of 16, 17–20 Usher, J.M. 40 vaccines, case study 17–18, 21–2 van der Aalst, W. 47, 62 van der Steen, M.P. 169 Van Dongen, B.F. 62 Van Maanen, J. 169 Varga, A. 248 Veblen, T.B. 27, 28, 29, 32, 248, 252 Versace 114 Vinodrai, T. 113 virtual simulation in experimentation 228–9 virtual simulation technology and innovation 223–5, 239–41 analytical perspective and method 228–30 empirical findings 230–39 innovation outcomes 232–9 opportunities provided 233–4 procedures’ leverage of opportunities 234–7 setting up 230–31 stability, effects of 237–9 use of 231–2 literature 226–7 new product development 227–8 organizational routines 225–6 Vromen, J.J. 267 Waddell, G. 107, 114 Wasserman, S. 60 Waterman, M.S. 74, 76 Watson, L. 107, 108, 110, 111, 117, 118, 124 Weick, K.E. 112, 132, 146, 162, 164, 171, 190, 250, 251, 254 Weijters, A.J.M.M. 47 Weissman, D. 35, 40 Wenger, E. 147, 214 Wenting, R. 104, 105 West, J. 206 West, M.A. 224

Index Wezel, F.C. 105, 125 Wheelwright, S.C. 224, 228, 234 Whetten, D.A. 146 Winter, S.G. 1, 19, 23, 29, 30, 31, 33, 37, 51, 68, 70, 71, 74, 79, 84, 90, 103, 111–12, 114, 166, 185, 215, 224, 225, 226–7, 228, 238, 248, 253, 267–8 Witt, U. 250 Wittgenstein, L. 178, 190 Wofford, J.C. 163 Wood, W. 40 Woods, C.R. 248 Woolgar, S. 191 workarounds 202, 204, 209 workflow data, use in routine structure analysis 47–8 discussion, limits and possibilities 61–4 antecedents of structures 63–4 combinatorics of organizational evolution 62–3 data, use of 61–2 sequential variety and evolution 63 structure identification 62 methodology 51–7

291

data, lexical variety 54–5 sequences listing 55–7 string matching distance 64–5 structure of routines 57–61 action networks 58–60 dynamic networks 60–61 network graphs and measures 58–60 sequential variety 57–8 theory 48–51 properties and patterns over time 50–51 workflow event logs 47–8, 52, 54, 61–2, 63 workflow systems 47, 53, 61 Worth, Charles Frederick 106–7, 108–9 Wynne, B. 217 Yin, R.K. 167, 229 Zander, U. 68 Zellmer-Bruhn, M.E. 112, 113 Zirpoli, F. 234 Ziv, J. 65 Znaniecki, F. 27 Zollo, M. 19, 185 Zucker, L.G. 36