It-Based Decision Tools for Item Processing Operations Management in Retail Banking

IT-Based Decision Tools For Item Processing Operations Management in Retail Banking

Charles J. Malmborg

IDEA GROUP PUBLISHING

IT-Based Decision Tools for Item Processing Operations Management

IDEA GROUP PUBLISHING

1331 E. Chocolate Avenue, Hershey PA 17033-1117, USA Tel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com

1

IT5600

IT-Based Decision Tools For Item Processing Operations Management in Retail Banking Charles J. Malmborg Rensselaer Polytechnic Institute, USA

EXECUTIVE SUMMARY Merit Bank is a multi-line financial services company with $75 billion in assets and approximately 1,000 retail branches distributed across 20 geographic divisions in 16 states. In 1999, Merit’s retail banking operations generated $2.1 billion of revenues and $1 billion in net income. Over the past decade, Merit’s aggressive acquisition and consolidation strategy in its retail and commercial banking divisions has significantly increased check processing volumes and motivated major investments in automated imaging technology and branch operations reporting systems. When these investments failed to reduce overall check processing costs, a consulting team was formed to define the breakthrough opportunities and best-in-class management practices needed to restructure under performing operations. By using updated scheduling criteria reflecting current business conditions and more fully exploiting imaging and branch reporting software, the consulting team successfully developed and implemented interfacing tools responsible for significant cost savings in check processing operations.

BACKGROUND Technology and business conditions in the retail banking industry are changing more rapidly today than at any time in recent history. Consolidations of independent financial institutions are occurring at an unprecedented rate. Some analysts predict that fewer than half of the financial institutions operating today will exist as independent enterprises by the end of this decade (Malmborg, 1999). Merit Bank is a 50-year-old financial institution with a history of innovation and successful adaptation to change. During the past five years, multi-line financial services institutions (MFSIs) such as Merit have been vying to become top players in a select group of high-growth regional markets as well as market share leaders in geographic areas where they have an established presence. Following a “build and buy” strategy of acquisitions and regional asset exchanges, Merit is seeking to develop profitable niches in various areas of consumer finance and investment services as well as in its Copyright © Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.

2 Malmborg

traditional fields of retail and commercial banking. Among these multiple business lines, retail operations have retained their importance to Merit by serving as an inexpensive source of operating funds (deposits) and a significant source of revenues. In the retail segment of the banking industry, Merit follows a strategy of building broad-based client relationships through continuous pursuit of best-in-class services. In a number of regional markets, inroads made by Merit in the retail area are pressuring smaller independent banks to reduce costs and re-focus on the specialized financial services for which they are best positioned. Many of these smaller institutions have been forced to follow Merit’s lead in rapidly introducing new products and services, improving customer access to funds and reducing operating costs while maintaining the “personal banking” relationships on which their businesses were founded. Following a merger with a major investment firm in the late 1990s, Merit accelerated the acquisition-based growth strategy that had been its hallmark since the early 1980s. By 1999, Merit transitioned into a leading MFSI with over $75 billion in assets and approximately 1,000 retail branches in 20 geographic divisions spread over 16 states. Throughout the period of rapid expansion, retail operations remained a priority responsible for over $2.1 billion of revenues (almost 40% of total revenues), and over $1 billion in net income during 1999 alone. Under pressure to maintain annual growth in retail operations in the 10% range, Merit launched a productivity improvement program during 1999 that targeted cost reductions of over $80 million. A significant portion of this was focused on leveraging investments in new information technology to achieve economies of scale in item processing operations to enable closings and consolidations of acquired item processing facilities. Item processing (IP) operations involve the retrieval of paper checks from retail points of transaction and associated processing, distribution and funds transfers to and from correspondent banks. Despite explosive growth in paperless financial transactions over the past two decades, strong consumer resistance has frustrated efforts by Merit and other MFSI’s to drastically reduce consumer dependence on paper checks. Subsequently, IP operations have remained a major back-room operation for Merit and most other full service financial institutions. Moreover, IP processing volumes have expanded rapidly at Merit as formerly independent local and regional banks have been acquired and Merit has sought to grow into a provider of IP services for smaller institutions. Merit’s IP operating policies in retail and commercial banking involve the coordinated management of document retrieval courier operations, check encoding and cash lettering. Like many financial institutions, Merit outsources courier services to vendors that provide contracts to financial institutions for a fixed number of vehicles and drivers to be directly scheduled by the management of the bank’s local IP operations centers. A predominant management objective in courier operations is to determine the routings and departure times for vehicles to retrieve checks and other paper transit that accumulate at retail branch locations during business hours. After pickup, the material is transported to the local IP operations center. As Merit expanded rapidly in the 1990s, its role as a vendor of IP services for smaller institutions grew substantially in many of its regional markets. This usually took the form of Merit’s local IP facilities taking in transit for various levels of processing from smaller banks, utilities, cable television companies, insurance companies and other businesses with high check volumes. The addition of external transit to Merit’s IP work stream significantly increased the variability of check processing volumes and has complicated the forecasting of workloads at IP operations centers. Since the large majority of check volumes through the early 1990s had been from Merit-operated retail branches, IP workloads had been considerably more forecastable based on historical business patterns in Merit’s retail branch networks. During most of this period, monthly transactions reports from retail branches were a useful guide in developing courier, encoder and cash letter schedules based on standard personnel scheduling models for minimizing operator idle time (Ftizsimmons and Fitzsimmons 1998; Starr 1996). The processing sequence at most of Merit’s IP operations centers is essentially the same. Transit dropped off by couriers is encoded, batched for proofing, sorted by destination bank and cash lettered

IT-Based Decision Tools for Item Processing Operations Management

3

to a federal reserve or commercial correspondent bank in the district where the transfer of funds is effected. Encoding of the large majority of individual items is accomplished at workstations where the destination bank, item amount and other data are recorded in magnetic characters on each check. Encoded items are transferred to high-speed reader/sorters that create a photographic record of each check and physically batch checks by the federal reserve or correspondent bank to which they are sent for collection. A cash letter is then generated for each batch which is sent, along with the paper checks, to the correspondent according to that institution’s schedule of cutoff times for off peak item handling fee discounts and one-day or two-day collected funds availability. For high volume correspondents, several cash letters can be sent on the same shift. The operating shift at most IP operations centers starts during the late morning and, at some facilities, can run as late as 2:00 a.m. on high-volume weekdays. When courier schedules result in late arrivals of transit to an IP operations center, more checks miss early cash letter deadlines resulting in higher item handling fees and delayed funds availability. Courier schedules with pickups that are too early may miss large accumulations of transit at branch offices that result from the daily pattern of retail business. These accumulations may then have to wait for pickup on the next business day. During the 1980s, Merit was a pioneer in successfully applying complex, integrated cost models for coordinating courier routing and scheduling, encoder scheduling and correspondent bank fee and availability schedules to support IP operations management, (see Malmborg and Lutley, 1989; Malmborg and Simons, 1989). However, these tools had not been effective during much of the 1990s as IP operations were expanded to serve more external clients and item work streams from acquired institutions were merged for processing in the same facilities. Concurrently, rising competition for market share among federal reserve and commercial correspondent banks was fostering new marketing strategies in the industry for check clearing services resulting in lower unit item handling charges. However, Merit had not been successful in exploiting this opportunity due to its inability to adapt IP operations to more frequent and rapid changes in the timing of off-peak item handling fee discounts and funds availability schedules offered by its correspondents. Recognizing the importance of operations to overall competitiveness, Merit management began to focus on alternative, more convention methods for improving the efficiency of document retrieval operations (Chase and Hayes, 1991; Haksever et al., 2000, Kotter et al., 1986).

SETTING THE STAGE Merit’s overall item processing costs had not decreased significantly during the 1990s despite two major investments in information technology upgrades meant to exploit higher operating volumes. Among these was the acquisition of automated encoding technology. Although it was hoped that these systems could be cost justified within a five-year period, the initial motivation in adopting the new technology was strategic. Merit sought to position itself as a leading IP services provider in most of the regional areas where it operated a retail branch network. The automated encoding systems were intended to displace manual encoding stations where part-time hourly operators would read and encode the face amount of individual checks for subsequent processing by reader/sorters. Early challenges in implementation of automated encoding technology included unacceptably high error rates in the imaging software tools used to interpret handwritten check amounts. During the 1990s, newer generations of imaging software had become increasingly robust. By the late 1990s, the automated encoding system was capable of processing the majority of Merit’s transit work stream with a level of speed and accuracy sufficient to relegate manual encoders to the supporting role of handling exception items. Although this improvement did reduce encoding labor costs and the direct dependence of encoder scheduling on courier scheduling, the investment in automated encoding software and hardware technology had not proven cost effective. IP operating costs had not been significantly reduced after the five-year start-up period as initially envisioned, even though imaging software and automatic encoders had significantly increased processing rates and potential encoding capacity.

4 Malmborg

Given rising check volumes and the de-coupling of courier scheduling and encoder scheduling, Merit executives believed that the investment could be better adapted to achieve meaningful cost savings in IP operations. Subsequently, IP was targeted as a priority area for restructuring during 1999 with an emphasis on improving business process flows on the part of the organization. (Anupindi et.al., 1999). A second information technology upgrade identified as a potential source of IP cost reductions was the Branch Operations Reporting and Information System (BORIS). BORIS is a proprietary software system developed by Merit to track retail transactions at the branch level. It was originally developed and rolled out to branches in 1998 as a general-purpose operations management and marketing support system to track business activity and build branch activity databases. A key feature available in BORIS is a code value assigned by tellers during customer transactions indicating the type of transaction processed. This parameter can be used to identify transactions resulting in checks for collection from banks outside the regional collection district and thereby identify transit accumulations sensitive to the timing of courier stops at a branch office. BORIS was not fully successful in achieving some of its original objectives, including support of courier scheduling and auditing of branch performance for timely proofing of transit bags for courier pick up. Delays in transit proofing by branch managers were believed to be one factor contributing to under-performing IP operations. Aside from its investments in information technology upgrades, Merit identified business trends that were favorable to achieving cost savings in IP operations. The most obvious was the consolidation of IP operations associated with the closure of processing centers operated by acquired banks as well as some of its own satellite encoding centers. While aggressive marketing of IP services to regional businesses was beginning to yield some increases in the utilization of the capacity of automated encoding systems, competition in the market for check clearing services was driving down unit handling fees charged by correspondent banks. The barrier to exploiting this development was that this competition also resulted in greater uncertainty and frequency of changes in funds availability and item handling fee schedules. Ironically, the same business changes promising lower IP operating costs were now contributing to Merit’s under performance. The previous system bottleneck associated with the coupling courier scheduling and encoder scheduling had just been shifted to a new bottleneck associated with the coupling of courier scheduling and cash letter dispatching.

CASE DESCRIPTION Challenged to turn around the situation in IP operations, Merit executives identified IP as a strategic opportunity for restructuring and made it a key element of an $80 million cost savings program implemented in 2000. The first step was to focus a team of internal consultants on a typical IP operations center to better understand current operations and define breakthrough opportunities. The key criterion for selecting a regional site was that it had to be representative of IP facilities serving important regional markets. Once the breakthrough opportunities were identified, the team would develop tools and management practices for best-in-class IP operating procedures that could be rolled out to other facilities. Prior to selecting a site, the project team examined the information systems used by Merit’s IP operations managers. The primary system for courier scheduling was the IP Decision Support System (IPDSS). The IPDSS module for courier scheduling was based on a relatively complex, integrated cost model that could be used for interactively optimizing courier schedules based on a static forecast of hourly check volumes by branch, and a cash letter dispatching plan assuming fixed funds availability and per-item handling fee schedules. The system was designed for interactive, simultaneous development of courier and cash lettering schedules that would remain static for a fixed period (weekly to monthly or longer), depending on the variability in item volumes for the branches included on a courier route. From the analysis of the IPDSS, the project team determined that its failure to effectively adapt to new technology and prevailing business conditions was attributable to changes in the relative

IT-Based Decision Tools for Item Processing Operations Management

5

importance of different capabilities for managing IP operations, particularly courier schedules. The frequent revision and dynamic development of good quality courier schedules for check retrieval was clearly more important in 2000 than construction of near optimal, static schedules based on fixed demands in a branch network with predictable item volumes. It was no longer feasible to directly couple check retrieval and processing with correspondent bank handling fee and funds availability schedules due to the instability of these schedules. Nonetheless, it was obvious that cost reductions from minimizing delays in item retrieval to achieve off-peak discounts in check clearing and maximization of funds availability were still fundamental to the success of IP operations. Whatever new innovations correspondent banks were introducing, the key to exploiting them was still the ability to expediently retrieve, process and dispatch checks on the same day that they are received at the branch office (Malmborg, 1999). Subsequently, the breakthrough opportunity identified by the project team focused on the networking of check accumulation data by branch from BORIS with the IPDSS, and then developing revised functionality for courier scheduling based on realistic cost and performance models. This innovation combined with the faster turnaround of transit associated with automated encoding systems would position Merit’s IP managers to adapt courier schedules to transactions flows in the branch network and rapidly changing cash letter schedules. Using interstitial software to exploit Web connectivity linking BORIS with the IPDSS would enable access to near real time data on check volumes by branch. These updates were a means to partially overcome the non-forecastability of check volumes arising from the merging of transit streams from different institutions that caused short term item volume forecasting to become highly accurate relative to long-term forecasting. This was opposite to the situation existing through the early 1990s when long-term forecasting was considerably more accurate and volumes coming into a given IP operations center were relatively stable. Knowing the location of the highest item accumulations in the branch network, IP managers could be more effective in dynamically scheduling couriers to minimize pick-up delays. High-speed automated encoding could then make checks available for cash lettering shortly after drop off at the IP facility. After detailed analysis and brainstorming by the consulting team, it was determined that the key to turning around Merit’s under performing IP operations was courier scheduling that would assure that checks received at branch offices were cash lettered to correspondent banks on the same day that they were received. At the same time, it could not be assumed that branch personnel could be required to wait for after-hours courier pickups. Therefore, the operational challenge for the team was to develop a general method for scheduling couriers to complete all stops before branch closing times while simultaneously minimizing the volume of checks that had to be held over for pick up on the next day. To develop and test a prototype implementation of this concept, the team selected the Merit IP operations center in Springfield, Massachusetts. The Springfield site was typical of many of Merit’s other facilities in several respects. Work flows were heavily influenced by external transit streams, highly variable and difficult to predict. Only about 40% of stop locations on courier routes were Merit branches using BORIS. In addition, Springfield was making heavy use of aggressive commercial banks based in New York City for check clearing as well as the regional federal reserve banks. Potentially, BORIS could support dynamic scheduling of couriers by enabling dispatchers to adjust courier routings and departure times on a more frequent basis. Therefore, Springfield was a good site to test the concept of courier scheduling with the simple goal of minimizing the number of checks held over for next-day pickup. This was a significant departure from the explicit linkage of courier schedules with item accumulation forecasts and cash letter deadlines that was programmed in the IPDSS. The advantage of the simple hold-over volume reduction concept was that it would avoid reliance on outdated assumptions concerning item processing rates, funds availability schedules and correspondent bank fee structures, while exploiting information provided through BORIS to address these criteria in an approximate fashion. High-speed automated encoding would also eliminate the added complication of offsetting courier drop off times to accommodate limitations in encoding capacity.

6 Malmborg

Another key feature of the Springfield facility was that it resembled other IP facilities by serving two distinct groups of stop locations. The first group included branches and other stop locations in the geographic core of the region. This area had the highest density of stop locations that were served by overlapping courier schedules requiring simultaneous, multiple-vehicle routings. After an investigation of current practices in Springfield, the team determined that the IPDSS was being applied with reasonable effectiveness for scheduling couriers serving this first group of stop locations. In this case, dispatchers took full manual control to direct the route building process while utilizing the IPDSS to estimate courier travel times and the spacing of courier pickups at individual stop locations (often by different couriers). The compressed timing of large check accumulations, the high density of stop locations and the generally earlier closing times for urban branches reduced this part of the scheduling process to one of simply avoiding obvious errors such as high mileage routings, early final pickups or redundant stops. Opportunities for improvement seemed limited. The second group of stop locations consisted of suburban and rural branches served by couriers dedicated to mutually exclusive clusters of stop locations. The problem of scheduling couriers serving this group of stop locations involved the determination of a starting time and sequence of stop locations to establish the departure times from branch offices that define a courier schedule. Like the branches in the first group, courier schedules for these branches were generated manually by dispatchers in Springfield using the interactive portions of the IPDSS. These schedules were then printed, implemented, and reviewed periodically when weekly or monthly summaries of check receipts, (generated using BORIS) were reported by stop location. The same courier schedule could remain in effect for several months or longer. The second group of stop locations involved more geographically dispersed branch offices but was still responsible for a high proportion of total item volume. Dispatchers seemed less effective in applying the interactive scheduling tools of the IPDSS to serve these stop locations and reported the process as being more difficult and time consuming. In particular, pick-up volumes were more sensitive to the schedule since only one or two stops were made at any given location on the same day, suburban branches usually had later closing times and item accumulations were spread more widely over the operating period. It was not possible to provide frequent stops at individual locations and simply avoiding redundant stops and inefficient routings was no guarantee against holding over large check accumulations. For these branches, the problem was one of routing couriers to minimize the number of checks accumulating at the branch offices between the final pickup of the day and the branch closing time. The prospect of more dynamic courier scheduling with revision of courier schedules on a weekly or even daily basis seemed more promising for the second group of locations. To achieve this, a method was needed to quickly and automatically generate efficient courier schedules based on near-real-time downloads of transactions data from BORIS. Focusing on the simple criterion of minimizing the number of checks held over, the team carefully analyzed the Springfield operation which included a total of 217 stop locations and an average of 14 courier schedules per day. For the suburban courier schedules, the operational issues boiled down to setting the time for dispatching each courier from the operations center and the sequence of branch locations for the final pickup. Estimating driving times between branches was not an issue of concern since this functionality was already available within the existing IPDSS. Driving routes, pick-up time allowances and an electronic map were used by the IPDSS database to compute travel time estimates with an accuracy of approximately +/- 4% based on weighting factors for route length, local area density and traffic patterns. Generating the software interfaces and coding to compute expected delay times using the IPDSS driving time models and transactions data from BORIS would be a simple matter for programmers once the basic scheduling method was established. Based on two fundamental relationships, the team concluded that development of a simple but reliable method for scheduling couriers to minimize the volume of checks held over at retail branches provided the most leverage for improving under-performing IP operations. First, minimizing hold-over volume was the key to utilizing the high item throughput capacity provided through automated

IT-Based Decision Tools for Item Processing Operations Management

7

encoding systems for just-in-time processing. These systems could easily encode a high volume of checks arriving relatively late in the day, leaving ample time for meeting most early cash letter deadlines. Second, maximizing the volume of checks cash lettered to correspondent banks on the same day of receipt at the branch was the prerequisite for exploiting item handling fee discounts and funds availability schedules offered by correspondent banks. Refining this scheduling concept and validating the effectiveness for the Springfield operation, and then rolling it out to other regional IP operations centers, had the potential to contribute directly and significantly to the $80 million cost savings target established for Merit’s retail division.

CURRENT CHALLENGES/PROBLEMS FACING THE ORGANIZATION There were four factors pressuring the team to achieve a quick success in turning around the Springfield IP operation. The first was a widespread recognition of the importance of retail operations to Merit’s balance sheet. As a source of low-cost operating funds and direct revenues, retail operations were a central part of Merit’s business strategy and were apt to remain so for the foreseeable future. Growing Merit’s market share as a provider of IP services was an integral part of this strategy. The articulation of this strategy as one aimed at achieving “best-in-class” performance in key areas made it all the more essential for the team to show results. The second factor was management’s emphasis on restructuring IP operations to achieve near term cost reductions. There was an expectation that a significant portion of the 1999 cost savings target of $80 million would be achieved through improvements in IP operations. The team’s project was among the most visible initiatives in this area. The third factor was a decreasing level of institutional patience with a lack of return on investments in information technology tools designed to improve the efficiency of retail operations. The high expectations associated with the rollout of BORIS a year earlier were contributing to a growing concern that the system could prove to be a costly failure. A significant cadre of managers that had supported its development were eager to see its reputation rehabilitated. The fourth factor was a problem created by the team itself. This was from overselling the concept that a simpler approach to managing IP operations would prove more effective than the traditional operations strategy. To build support and buy-in for its plan, the team had to convince management stakeholders that scheduling couriers to minimize hold-over volume could turn around under performing IP operations. This tended to oversimplify the problem by under-emphasizing other changes needed in the way that Merit’s IP operations were managed. The result was a perception that the problem was considerably easier than it actually was and that results should be forthcoming in the short run. There were also a number of implementation-related and technical problems associated with the primary user community of courier dispatchers. The chief technical problem was designing a scheduling method that would work effectively for IP centers throughout the Merit system where there was wide variation in the urban/suburban mix of courier schedules, the average number of stops on courier routings and other factors influencing the scheduling problem. The method had to feature an initial schedule development routine that was mostly independent of manual user inputs and then provide an efficient interface for schedule adjustment that exploited the domain knowledge of dispatchers. Therefore, a clear understanding of the core scheduling logic of the system was seen as a key to gaining credibility and ultimately acceptance among the dispatcher community. Recognizing the time pressure under which most dispatchers were working, the team had to sell the concept that the new tool would produce immediate results by better exploiting the computational power of the IPDSS. Finally, there was the problem of reliance on the somewhat discredited BORIS as a key source of inputs. The user community was skeptical of the idea of adapting databases within BORIS to support the IPDSS. The team had to demonstrate that the use of this data would produce schedules that were of sufficient quality with respect to holdover times to either replace or at least benchmark existing schedules.

8 Malmborg

Recognizing these implementation issues, the technical approach developed by the team was to use Monte Carlo simulation for generating courier schedules within starting time windows provided by local dispatchers. The technique used “breadth” and “depth” parameters where the breadth parameter corresponded to the number of different branch sequences to generate for a particular courier schedule and the depth parameter represented the extent to which sequences were “improved” prior to fixing a schedule. For example, for a breadth parameter of 50 and a depth parameter of 1,000, the procedure would generate 50 random routings (branch sequences), and then evaluate the expected volume of checks held over for each sequence using data from BORIS and the travel time calculation functions of the IPDSS. Any sequence that was infeasible (i.e., could not be completed before a given deadline) would be discarded and replaced with another one until 50 feasible random routings were obtained. For each of these 50 random routings, the improvement step would be applied 1000 times. The improvement step involved exchanging the locations of two randomly selected stops in the routing, and then reevaluating the corresponding hold-over volume. If the exchange yielded an improvement, i.e., a lower hold over volume and still met feasibility criteria, it was retained. Otherwise it was discarded. The IPDSS could be used to print any dispatcher-specified subset of the 50 improved schedules, and its existing schedule editing functions could be applied to evaluate and implement dispatcher-defined changes for any schedule prior to printing. The team’s concept was therefore to use real-time transactions data from BORIS and the schedule building and editing utilities of the IPDSS. The method was practical, simple and would scaleup efficiently for courier routes with a very large number of stop locations. It can be illustrated using the sample data set with the check accumulation rates and courier driving times summarized in Table 1. The sample problem has eight branches that operate over the same eight-hour period each day and the daily check volume associated with each branch is obtained by totaling the volume in each hour over the eight hour operating period to obtain: Branch Office: B1 B2 B3 B4 B5 B6 B7 B8 Daily Total (checks): 913 1364 1862 2271 2457 2479 2157 1952 where branch office j is designated as Bj. These values are obtained by multiplying the estimated number of checks accumulating per minute in each hour of operation by 60 minutes per hour, e.g., 913 = 60(2.29+1.21+2.10+1.65+2.03+2.94+1.48+1.51). With a breadth parameter of two and a depth parameter of five, a total of 10 schedules would be examined To illustrate, suppose the following two sequences that start and finish at the item processing operations center (IPOC) define the two randomly generated branch sequences: Sequence 1:

IPOC

B2

B7

B6

B3

B5

B1

B4

B8

IPOC

Sequence 2:

IPOC

B3

B6

B8

B1

B7

B2

B5

B4

IPOC

The first step would be to evaluate each of these sequences. The schedule driving times are obtained using the driving time data from Table 1 and the schedules are generated by setting the schedule starting time to allow exactly enough time for a courier to complete the schedule. That is, the schedule starting time is set to allow the courier to arrive at the IPOC at the end of the eight- hour (480minute) operating period. For example, the schedule associated with sequence 1 starts at minute 235 in order to complete at minute 480. The schedule associated with sequence 2 starts at minute 222 in order to complete at minute 480. For each stop on each schedule, the pickup volume is computed by tabulating the number of checks that arrive at the branch office prior to the pickup time. The hold-over volume is then obtained as the difference between the total daily volume and the pickup volume. The calculations for sequences 1 and 2 are summarized below: Sequence 1:

IPOC

B2

B7

B6

B3

B5

B1

B4

B8

Stop Time: Pickup Vol.: Hold Over:

235 -

280 843 69

315 739 625

351 1571 291

394 2033 238

428 2225 232

434 1792 687

438 1264 893

464 480 1882 70

IPOC

IT-Based Decision Tools for Item Processing Operations Management

9

Sequence 2:

IPOC

B3

B6

B8

B1

B7

B2

B5

B4

Stop Time: Pickup Vol.:

222 -

248 724

291 1137

337 904

357 2129

366 2381

401 1326

455 1615

463 480 1300 -

IPOC

Hold Over:

-

188

227

958

142

76

1153

542

652

-

Thus, total hold-over volumes of 3106 and 3938 checks result from schedules 1 and 2 respectively. To perform the second, “improvement” step in the scheduling procedure, two branches in a sequence are randomly interchanged. If the hold-over volume is reduced, an interchange is retained otherwise it is discarded. The results below summarize the improvement steps for each of the two initial sequences for a depth parameter of five: Initial Sequence 1: IPOC-B2-B7-B6-B3-B5-B1-B4-B8-IPOC (hold over volume of 3106) Adjusted Sequence: IPOC-B2-B7-B6-B1-B5-B3-B4-B8-IPOC IPOC-B5-B7-B6-B3-B2-B1-B4-B8-IPOC IPOC-B2-B5-B6-B3-B7-B1-B4-B8-IPOC IPOC-B2-B5-B4-B3-B7-B1-B6-B8-IPOC IPOC-B2-B5-B6-B3-B7-B8-B4-B1-IPOC

Interchange: B1 & B3 B2 & B5 B7 & B5 B4 & B6 B1 & B8

Holdover: 3696>3106 4205>3106 3007<3106 3801>3007 3090>3007

Retain/Reject Reject Reject Retain Reject Reject

Initial Sequence 2: IPOC-B3-B6-B8-B1-B7-B2-B5-B4-IPOC (hold over volume of 3938) Adjusted Sequence: IPOC-B3-B6-B8-B1-B4-B2-B5-B7-IPOC IPOC-B2-B6-B8-B1-B4-B3-B5-B7-IPOC IPOC-B2-B6-B8-B5-B4-B3-B1-B7-IPOC IPOC-B2-B8-B6-B5-B4-B3-B1-B7-IPOC IPOC-B2-B8-B6-B5-B1-B3-B4-B7-IPOC

Interchange: B4 & B7 B2 & B3 B1 & B5 B6 & B8 B1 & B4

Holdover: 4232>3938 3440<3938 3430<3440 3281<3430 3136<3281

Retain/Reject Reject Retain Retain Retain Retain

Thus, the best schedule obtained using a breadth parameter of two and a depth parameter of five is: IPOC-B2-B5-B6-B3-B7-B1-B4-B8-IPOC, with a total of 3007 checks held over.

APPENDIX Table 1: Summary of Data For Test Problem 1 Estimated Driving Times (in minutes) Proc. Branch Office Ctr.: #1: #2: #3: #4: #5: #6: #7: #8: Proc. Ctr. - 17 45 26 29 17 72 13 16 Branch #1 17 - 40 27 4 6 26 9 20 Branch #2 45 40 - 29 54 48 106 35 30 Branch #3 26 27 29 - 35 34 43 20 10 Branch #4 29 4 54 35 - 8 40 12 26 Branch #5 17 6 48 34 8 - 37 17 28 Branch #6 72 26 106 43 40 37 - 36 46 Branch #7 13 9 35 20 12 17 36 - 12 Branch #8 16 20 30 10 26 28 46 12 Item Accumulation Averages Per Minute for Each Hour of Operation Branch No: 1: 2: 3: 4: 5: 6: 7: 8: Hour 1 2.29 4.29 5.31 8.55 8.11 10.04 7.52 6.72 Hour 2 1.21 1.48 2.98 3.51 4.08 3.34 2.83 2.32 Hour 3 2.10 2.17 2.03 2.24 2.41 2.33 2.16 2.19 Hour 4 1.65 2.96 4.43 4.40 5.03 4.65 4.84 4.30 Hour 5 2.03 2.12 2.34 2.54 2.14 2.09 1.93 2.27 Hour 6 2.94 4.35 7.00 6.72 10.20 8.77 7.14 6.27 Hour 7 1.48 2.31 3.70 4.22 4.52 4.67 4.92 4.08 Hour 8 1.51 3.05 3.24 5.67 4.46 5.47 4.61 4.38

10 Malmborg

FURTHER READING Evans J.R. & Olson D.L. (2000). Introduction to Simulation and Risk Analysis. Upper Saddle River, NJ: Prentice Hall. Fishman G.S. (1996). Monte Carlo Concepts, Algorithms and Applications. New York, NY: SpringerVerlag New York, Inc. Tapiero C.S. (1998). Applied Stochastic Models and Control for Finance and Insurance. Hingam, MA: Kluwer. Crainic T.G. & LaPorte G. (1998). Fleet Management and Logistics. Hingam, MA: Kluwer. Golden B.L. & Assad A.A. (1988). Vehicle Routing: Methods and Studies. Amsterdam, Netherlands: Elsevier Science Publishers.

REFERENCES Anupindi, R., Chopra S., Deshmukh D., Van Mieghem J, & Zemel E. (1999). Managing Business Process Flows. Upper Saddle River, NJ: Prentice Hall. Chase R.B. & Hayes R.H. (1991). Operations’ Role in Service Firm Competitiveness, Sloan Management Review, 33(1), 15-26. Fitzsimmons J.A. & Fitzsimmons M.J. (1998). Service Management: Operations, Strategy and Information Technology. Boston, MA: Irwin Mc-Graw-Hill. Haksever C., Render B., Russell R.S., & Murdick R.G. (2000). Service Management and Operations. Upper Saddle River, NJ: Prentice Hall. Kotter J.P., Schlesinger L.A. & Sathe V. (1986). Management of Service Operations. Homewood, IL: Irwin. Malmborg C.J., Lutley R. (1989). A PC Based System for Financial Transit Retrieval Operations. Industrial Engineering. 21(12), 30-36. Malmborg C.J., Simons G.R. (1989). Integrating Logistical and Processing Functions Through Mathematical Modelling. Applied Mathematical Modelling. 13(6), 357-364. Malmborg C.J. (1999). Current Modeling Practices in Bank Courier Scheduling. Applied Mathematical Modelling, 24(4), 315-325. Starr M.K. (1996). Operations Management: A Systems Approach. New York, NY: International Thompson Publishing Company.

BIOGRAPHICAL SKETCH Charles J. Malmborg is Professor and Acting Chair in the Department of Decision Sciences and Engineering Systems and Professor of Information Technology at Rensselaer Polytechnic Institute in Troy, New York. His research interests are in material flow logistics, facility design and operations management. He teaches courses in Rensselaer’s information technology, industrial engineering, management of technology distance learning and executive MBA programs. He is author or co-author of more than 60 refereed technical papers in leading international journals and has served as Project Director and Principal Investigator on over $2 million in sponsored research and educational projects.