PRODUCTION AND OPERATIONS MANAGEMENT POMS Vol. 19, No. 6, November–December 2010, pp. 633–664 DOI 10.3401/poms.1080.01163 ISSN 1059-1478|EISSN 1937-5956|10|1906|0633 r 2010 Production and Operations Management Society

Operations in Financial Services—An Overview

Emmanuel D. (Manos) Hatzakis Goldman Sachs Asset Management, Goldman, Sachs & Co., New York, New York 10282-2198, USA, [email protected]

Suresh K. Nair Department of Operations and Information Management, School of Business, University of Connecticut, Storrs, Connecticut 06269-1041, USA, [email protected]

Michael L. Pinedo Department of Information, Operations and Management Science, Leonard N. Stern School of Business, New York University, New York, New York 10012-1106, USA, [email protected]

e provide an overview of the state of the art in research on operations in financial services. We start by highlighting a W number of specific operational features that differentiate financial services from other service industries, and discuss how these features affect the modeling of financial services. We then consider in more detail the various different research areas in financial services, namely systems design, performance analysis and productivity, forecasting, inventory and cash management, waiting line analysis for capacity planning, personnel scheduling, operational risk management, and pricing and revenue management. In the last section, we describe the most promising research directions for the near future. Key words: financial services; banking; asset management; processes; operations History: Received: November 2009; Accepted: July 2010 by Kalyan Singhal, after 2 revisions.

1. Introduction insurance (other than health), credit cards, mortgage banking, brokerage, investment advisory, and asset Over the past two decades, research in service oper- management (mutual funds, hedge funds, etc.). ations has gained a significant amount of attention. Special issues of Production and Operations Management have focused on services in general (Apte et al. 2008), 1.1. Importance of Financial Services and various researchers have presented unified theories Financial services firms are an important part of the (Sampson and Froehle 2006), research agendas (Roth and service sector in an economy that has been growing Menor 2003), literature surveys (Smith et al. 2007), rapidly over the past few decades. These firms pri- strategy ideas (Voss et al. 2008), and have discussed marily deal with originating or facilitating financial the merits of studying service science as a new discipline transactions. The transactions include creation, liqui- (Spohrer and Maglio 2008). A few books and a special dation, transfer of ownership, and servicing or issue of Management Science have focused on the oper- management of financial assets; they could involve ational issues in financial services in particular (see raising funds by taking deposits or issuing securities, Harker and Zenios 1999, 2000, Melnick et al. 2000). making loans, keeping assets in custody or trust, or However, financial services have still been given scant managing them to generate return, pooling of risk by attention in much of the literature relative to other underwriting insurance and annuities, or providing service industries such as transportation, health care, specialized services to facilitate these transactions. entertainment, and hospitality. The dilution of focus, by Services is a large category that encompasses firms concentrating on more general distinguishing features as diverse as retail establishments, transportation does not do justice to financial services where some of firms, educational institutions, consulting, informa- these characteristics are not central. (The more general tion, legal, taxation, and other professional, real estate, features that are typically being considered include and healthcare. Even within financial services, there is intangibility, heterogeneity, contemporaneous produc- a wide variety of firms which are characterized by tion and consumption, perishability of capacity, waiting unique production processes and specialized skills. lines (rather than inventories), and customer participa- The processes and skills required for banking are tion in the service delivery.) quite distinct from solicitations for credit cards, ac- In this overview, we mean by financial services quisitions of new insurance accounts, or the handling primarily firms in retail banking, commercial lending, of equity dividends and proxy voting, for example. 633 Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 634 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

Table 1 Employment within Financial Services Table 2 A sample of Financial Service Operations Job Titles, Responsi- bilities, and Products Total employees Titles Products NAICS code Title within financial services Number % Vice President, Operations, Opera- Premiums, Claims, Refunds, Cash 5211 Monetary authorities-central bank 21,510 0.4 tions Manager, Financial Operations flow and treasury management, (Federal Reserve banks, etc.) Supervisor, Foreclosure and Bank- Customer statements, Loan servi- 5221 Depository credit intermediation 1,816,300 30.7 ruptcy Operations Manager, Risk cing and support, Trade (banks, credit unions, etc.) Operations Team Manager, Team confirmations, Reconciliations, Tax 5222 Non-depository credit intermediation 659,930 11.2 Manager Ops Control-Fixed Income, reporting, Security settlements, (credit cards, mortgage lending, etc.) National Director Operations, Hedge Mortgages Fund Operations Specialist, VP/ 5223 Activities related to credit intermediation 294,910 5.0 Director Of Operations (brokers for lending) 5231 Securities and commodity contracts 516,010 8.7 Nature of work/responsibility intermediation and brokerage 5232 Securities and commodity exchanges 8,010 0.1 Brokerage operations, Improve customer service, resolves customer issues, Review security pricing, Vendor support, Authorize net settlement, Hand-off of 5239 Other financial investment activities 344,950 5.8 data, ensure data integrity, Verifying transactions, Tracking missing transactions, (mutual funds, etc.) Leverage technology, Maintain ops controls, update policies, procedures, Back- 5241 Insurance carriers 1,258,050 21.3 office support, Understand regulations, Ensure compliance, Attain profit and 5242 Agencies, brokerages, and other 907,880 15.3 revenue benchmarks, Reduce risk, Improve quality, Six sigma, Operational insurance-related activities processing efficiency, Problem solving, Ensure best practices, Streamline 5251 Insurance and employee benefit funds 47,730 0.8 activities, Manage key expenses, Work management tools, Monitors work flow, Production/testing 5259 Other investment pools and funds 41,190 0.7 5,916,470 100.0 Source: Monster.com jobs listings during the week of March 8, 2009.

Source: Bureau of Labor Statistics, stat.bls.gov/oes/home.htm, May 2008. the services economy, we wish to explore whether Even though services account for about 84% of the operations in financial services are indeed unique, or total employment in the economy, only about 4% of share several characteristics with services in general. this workforce is employed in financial services. This That is the motivation for this special issue. might come as a surprise to some because financial services transactions in one form or another are so 1.2. Distinctive Characteristics of Operations in ubiquitous in our lives. Not surprisingly, however, the Financial Services number of financial services firms is about 7% of the There are several unique operational characteristics total non-farm firms and contributes about 13% of to- that are specific to the financial services industry and tal non-farm sales. Only wholesale trade has a similar that have not been given sufficient attention in the employment and number of firms with a larger con- general treatment of services in the extant literature. tribution to sales. We list below a number of these unique operational Table 1 provides employment information for the characteristics and elaborate on them in what follows: sub-codes within financial services. As can be seen, retail banks, insurance companies, and insurance bro-  Fungible products with an extensive use of kers together employ about two-thirds of the financial technology services workforce. A cursory look at the table gives a  High volumes and heterogeneity of clients sense of the diversity of the services sector. Clearly  Repeated service encounters operations management problems and approaches  Long-term contractual relationships between used to solve them have to be customized for partic- customers and firms ular types of services—we already know that what  Customers’ sense of well-being closely inter- works for manufacturing may not work for services, twined with services but by looking at Table 1 we can also realize that what  Use of intermediaries works for retail trade or recreational services may not  Convergence of operations, finance, and marketing. work for financial services. A quick glance through Monster.com’s job openings for operations managers 1.2.1. Fungible Products with an Extensive Use in financial service firms shows a wide variety of ti- of Technology. One obvious difference between tles, responsibilities, and ‘‘products’’ related to such operations in financial services and operations in jobs. This is shown in Table 2, and gives a sense of the manufacturing and in other service industries is that wide swath of topics that could be covered in aca- the ‘‘widgets’’ in financial services are money, or related demic research on financial services operations. As financial instruments. As there is a declining use of the financial services are such an important segment of physical vestiges of money such as coins, currency, Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 635 bond, or stock certificates, much of the transactions are a client could wipe out all the profit from the client’s in the form of bits and bytes. Thus inventory is fungible savings account. Very little research in service operations and can be transported, broken up, and reconstituted management has focused on this issue. New customers (facilitating securitization, e.g.) in malleable ways that constitute another major group that is more likely to are simply not possible in manufacturing or in other make calls with billing questions or inquiries regarding service industries (see, e.g., the recharacterization of their statements. Should billing and statementing to new bank reserves in Nair and Anderson 2008). customers be handled differently, perhaps with more The increased use of online transactions (in broker- care, than to existing customers? Obviously, if this ages, credit card payments, retail banking, and retire- observation is true, differential handling can reduce the ment accounts, e.g.) are forcing fundamental changes in traffic to call centers. Existing call center research usually the way operations managers think about capacity is- assumes the call volume to be a given, for the most part, sues (for statement mailing and remittance processing, and the focus is on ‘‘managing’’ the traffic. This is akin to or for transfer of ownership in securities, e.g.). The fact traditional manufacturing where it was assumed that that adoption of online transactions is still growing and large setup times were a given, and a good way to has not yet matured and leveled off makes capacity ‘‘manage’’ would be to use an optimal batch size. One planning a big challenge. Yet we are aware of very little learns to live with such a constraint. Not until just in research that would help managers deal with this issue. time (JIT) manufacturing came along did managers question why setup times were large and what could be 1.2.2. High Volumes and Heterogeneity of Clients. done to reduce them. Financial services are characterized by very high At one credit card company, operations managers volumes of customers and transactions. Furthermore, struggled for years to cope with volatile demand in customers are not all alike. In many firms, a small bill printing, mailing, remittance processing, and call fraction of the customers generate most of the profits, center operations. Daily volumes could fluctuate giving the firms an incentive to view them differently easily between half a million and one and a half and provide differential treatment, given the firms’ million pieces of mail in remittance processing, and limited resources. For example, high net worth managers were reconciled to high overtimes and idle individuals may be treated differently by asset times because they felt they had no control over management firms; banking clients who keep high when customers mailed in their checks, and reduc- balances in checking accounts and transact heavily may ing float was important. Call volumes at the call be handled differently from depositors who keep almost centers were similarly volatile, as were volumes in all their funds in savings accounts and certificates of the bill printing and mailing operations. This situa- deposits (CDs) and transact minimally; revolvers (i.e., tion continued until someone recognized that all customers who carry over balances from one month to these problems were interconnected and to a large thenext)mayberegardeddifferentlyfromtransactors extent within the control of the firm. As it happens, (customerswhodonotrevolvebalances)byacredit the portfolio of current customers is distributed into card firm. In most non-financial services, because of about 25 cycles, one for each working day of the a limited number and sporadic interactions with month. For example, customers in the 17th cycle are customers (e.g., in restaurants and amusement parks), billed on the 17th working day of the month. Care is one customer is considered, for the most part, similar to taken to ensure that customers in the same zip code the next one in terms of margins and attention required. are put in the same cycle so volume-mailing dis- counts from the US postal service can be obtained, 1.2.3. Repeated Service Encounters. In contrast to and the cycles are level loaded. However, this allo- other service industries, where research typically cation to cycles was carried out several years back focuses on a single encounter (‘‘the moment of and over time some customers had closed their truth,’’ ‘‘when the rubber hits the road’’), financial accounts while new customers had been added to services are characterized by repeated service existing cycles, resulting in large differences in encounters or potential encounters between the firm the numbers of customers in the various cycles, and its customers due to regular monthly statements, and in a wide variability in the printing and mailing year-end statements, buy/sell transactions, insurance of monthly statements. On the remittance side, an claims, money transfers, etc. Anecdotal evidence from analysis found that there was less randomness in the brokerage and investment advisory industry customer payment behavior than one would expect. suggests that clients with low asset balances and There were broadly four cohorts of customers: transaction volumes contribute the least to firm revenue one that sent in payments on receipt of the state- and the most to operational cost through calls for ment, a second that mailed checks based on due customer service. One online bank discouraged calls to date, a third that acted based on salary payment customer service because it found that just a few calls by date, and a fourth that acted randomly. The first two Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 636 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society cohorts, the largest ones, were in fact dependent on activities (money the firm pays out but is owed the cycles that the firm had set up many years back. to it by other carriers). Timely intervention can Similarly, billing calls were also heaviest soon after avoid expiry of opportunities to collect dollars owed, the statement was received by the customer, again and more attention could be paid to even small traffic that was determined by the cycles created by opportunities. There is an extensive literature in risk the firm. and insurance journals on scoring for fraud pre- If the cycles could now be level loaded, many of vention and detection, but leveraging that informa- these problems would disappear (similar to what tion in the claims process can benefit from an happened in manufacturing when setup times were operations perspective. dramatically reduced thereby enabling lean opera- Another example from the insurance industry tions). But there was a problem—the firm needed to concerns worker’s compensation claims, where the inform each customer if their cycles were moved, for process for handling workplace injury can have good reason because customers needed to plan their long tails spanning several years before the claim finances. However, this notification was not neces- is closed. The process is complicated with inter- sary if the move was to be within Æ 3 days from actions between the worker, the employer, medical their current cycle. An optimization model found practitioners, hospitals, state authorities, and law- that this constrained move was sufficient to take care yers (both on the staff of the insurance firm and of the vast majority of moves that were initially panel counsel, i.e., lawyers who are hired on an thought to be necessary. ad hoc basis). There are several opportunities This example illustrates how stepping back and here (Jewell 1974) to speed up the process (and taking a broader view of the situation and collabo- speed up the workers’ return to work, which is in rating across processes can have a major impact on the insurance firm’s interest), reduce costs, detect financial service operations, something that is lack- fraud, ensure that review triggers are not over- ing in the current literature. looked, increase the utilization of staff counsel compared with panel counsel by better scheduling 1.2.4. Long-Term Contractual Relationships Between of appearances for hearings, etc. We are unaware Customers and Firms. Connected to the previous of any recent operations management literature in characteristic of repeat encounters is the recognition this area. that, unlike in other services, in financial services the firm and the customer have a relatively long-term 1.2.5. Customers’ Sense of Well-Being Closely contractual arrangement. However, technology and in- Intertwined with Services. Along with the ease of formation availability makes comparison shopping manipulating the putty at the core of the financial easy, resulting in easy switching between firms, and services process comes the responsibility of working thereforehighattrition.Thislossofcustomersmakesthe with something that is so close to the customer’s sense acquisition process very important to the continued of well-being and worth. Poor operations manage- growth and profitability of the firm. Similarly, loyalty ment that results in delays, quality issues, or sloppi- programs (such as rewards and balance transfer ness can and will attract regulatory scrutiny and programs in the credit card business) are important to unfavorable publicity, and will generate immediate stanch the bleeding. The design and execution of these rebukes from the customer in the form of calls, programs are based on complicated processes that need complaints, and because the account can be easily to consider risks, costs, redemptions, incremental sales, moved around, customer attrition. At least two factors scheduling and sequencing of offers, etc. Researchers in make the detection of errors due to operational faults financial services operations, by not making their and their exposure to the clients relatively easy in presence felt in these areas, are missing the boat with financial services: regard to issues that are the most important (‘‘must do’’ (i) the amount, frequency, and detail of communi- activities) for the firm, and may be paying instead too cation and disclosure as required by regulation, much attention to relatively mundane and low-impact and issues (‘‘good to do’’ activities). (ii) the clients’ heightened propensity and incentive Just as the above processes aim at increasing rev- to check for error in something so closely linked enue, there may be other processes that are put in to their livelihood and sense of security. place to reduce unnecessary costs. In the insurance business, for example, the claims processes may pri- Because of the above, tolerance for error is signifi- marily revolve around a call center, which has cantly lower than in other industries. For example, attracted sufficient attention in the literature as we faulty processes resulting in incorrect calculations of will see later. But unnecessary costs can be reduced interest amounts in savings, mortgage loan, or credit by fraud prevention and detection, and subrogation card accounts, or in inappropriate handling of stock Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 637 dividend payments, become obvious immediately acquisition process in full-service retail brokerage and after monthly statements are sent out. investment advisory firms begins at the corporate level, The above customer service issues are distinct where it draws resources from marketing, strategy, from the perceived quality of the performance of in- information technology, and operations, and is ulti- dividual customer brokerage portfolios, retirement mately implemented through the sales force of brokers/ accounts, annuities, mutual funds, and interest accrued financial advisors. Customer acquisition at a credit card in retail banking. With the plethora of information company is a competitive differentiator and a complex available comparing a customer’s firm with others in process focused on direct mail campaigns. At many the same space, moving an account to the competition large firms, the budgets for direct mail run into several is only a few clicks away. Eventhoughperformance hundred million dollars annually. By focusing on the may depend on the economy, the stock market, invest- billing mailroom, collections, call centers, and billing ment research, and fund manager performance, call centers, researchers in service operations are work- recognizing the costs (Schneider 2010) and capacity ing on a problem akin to quality inspectors at the end of issues (the increased transactions during market tur- the production line—by then it is too late, the volumes bulence, e.g.) are important operations management of mail and calls are baked in during the mailing concerns that have received little attention. campaign creation, while their skills could have made themailingsmoreeffectiveandtargeted(giventhe 1.2.6. Use of Intermediaries. This is an important minuscule response rates), resulting in fewer delin- aspect of the financial services industry. In some cases quent accounts (requiring fewer outbound collection a direct-to-consumer approach is used (credit cards); calls), and perhaps also fewer billing calls to inbound in other cases most of the customer facing work is call centers. done by intermediaries (financial advisors, insurance At a more sophisticated level, very few credit card agents, annuity sales through banks, etc.); and in still firms use contact history in mailing solicitations, other cases the firm’s employees and its agents have which may result in repeated mailings to chronic to collaborate with one another (insurance). Working non-responders. The managers developing campaign through an intermediary entails a set of issues not strategies may not have the analytical background that normally seen in other services that function without a researcher in service operations can bring to bear, intermediaries. For example, financial product and and the cycle time for campaign creation is typically service design and delivery get filtered through the so long and complicated that much attention gets prism of what the agent feels is in his or her own best expended on scoring for credit and response, file interest. At times, the relationship between the firm transfers from credit bureaus and data vendors, scrub- and the intermediary is not exclusive, hereby adding a bing of data, etc. These complicated processes leave layer of complexity because the customer may choose little time to incorporate experience from a previous between products from competing firms. Therefore, mailing because a reading of the results of that mailing what gets planned in the corporate offices of the takes time (prospective customers may not respond financial services firms and what is seen by the immediately, even if they do respond), and file struc- customer may be quite different. The operations tures may not have been designed to carry information management literature, to our knowledge, has not about previous contacts and the response to them. paid attention to product and service design in such Another indication of this convergence is that the situations, because the implications on customer majority of the undergraduate and MBA hiring at an lifetime interactions with the firm go much beyond investment bank in the greater New York City area initial pricing, product features, and the inventory of from one of the region’s business schools is in the brochures left with the agents. COO’s operation, whether the student concentra- tions are in finance, marketing, information systems, 1.2.7. Convergence of Operations, Finance, and or operations. Marketing. There is probably no other industry where The foregoing does not imply that no significant this convergence is more pronounced. These functions work has been done in the research on financial are supported and enabled by a healthy dose of service operations, just that areas of work have had a statistics, technology, and optimization. By focusing narrow focus. The purpose of this article and this only on back-office operations such as call centers, special issue is to begin to expand that focus and researchers in service operations are leaving a lot on the encourage research in neglected or emerging areas in table. There is very little research in the service financial service operations. We will survey existing operations literature that leverages this convergence, research next, not only where the attention is only on which requires a choreography, as described by Voss financial service operations, but also where research et al. (2008), who put it in a more limited context of in service industries in general has substantial appli- operations and marketing. For example, the client cation in financial service operations. In Appendix Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 638 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

A, we provide an overview of the various operations ing by creating front-office/back-office functions processes in financial services and highlight the (Sampson and Froehle 2006, Shostack 1984). Shostack ones that have been addressed in operations man- also pioneered the use of service blueprinting for agement literature. identifying fail points where the firm may face quality This survey paper is organized as follows. Sections problems. She illustrated this methodology for a dis- 2 through 9 go over eight research directions that are count brokerage and correctly identified that many of of interest from an operations management point of the operational processes are not seen by the customer; view. The first couple of sections consider the more she then focused on the telephone communication step, general research topics, whereas the later sections go the only one with client contact. This focus on client into more specific topics and more narrowly oriented contact tasks, whether in the front office or in the back research areas. Section 2 focuses on process and sys- office, is widespread in services research in general and tem design in financial services, while section 3 in research on financial services operations in particular. considers performance measurements and analysis. One reason may be that service researchers have found Section 4 deals with forecasting, because forecasting it necessary to motivate their work by differentiating plays a major role in virtually every segment of the services from products (whether it is service marketing financial services industry. The next section focuses vs. product marketing, or service design vs. product on cash and liquidity management; this section re- design), and client contact is an obvious differentiator. lates cash management to classic inventory theory. From the outset, it has been clear that service Sections 6 and 7 deal with waiting line management processes are subject to a significant amount of ran- and personnel scheduling in retail banking and in domness from various sources. Frei (2006) discusses call centers. Even though these two topics are strongly the various sources of randomness in service processes dependent on one another, they are treated separately; and how firms react to them in the design of their the reason being that the techniques required for deal- services. She identifies five types of variability— ing with each one of these two topics happen to be customer arrival variability, request variability, cus- quite different from one another. Section 8 focuses tomer capability variability, customer effort variability, on operational risk in financial services. This area and customer preference variability. She states that has become very important over the last decade and firms design services to factor in this variability by this section describes how this area relates to other trying either to accommodate the variability at a higher research areas in operations management, such as to- cost (cross training of employees, increased automa- tal quality management (TQM). Section 9 considers tion, variable staffing) or to reduce the variability with product pricing and revenue management issues. The a view to increasing efficiency rather than cost (off last section, section 10, presents our conclusions and peak pricing, standard option packages, combo meals). discusses future research directions. 2.2. Focus on Single Encounters 2. Financial Services System Design Much of the services literature, however, focuses on Service systems design has attracted quite a bit of single service encounters, which are common in ser- attention in the academic literature. It is clear that vices such as fast food. Even if a customer repeatedly service design has to be as rigorous an activity as visits the same restaurant, there is not the kind of product design, because the customer experiences the stickiness to the relationship as can be found in service first hand, much like a product, and comes financial services. Retail banking seems to have away with impressions regarding the quality of ser- attracted the most attention among financial services vice. Although the quality of service delivery depends with respect to service design, but here again the on a number of factors, such as associate training, focus is on disparate single visits to the branch or technology, traffic, neighborhood customer profile, Automated Teller Machine (ATM), rather than as part access to the service (channel access), and quality of of a life cycle of firm–customer interactions. Other resource inputs, the service experience gets baked into than meeting the branch manager when opening an the process at the time of the service design itself, and account, there is usually no other recognition of the therefore a proper service design is fundamental to stage of the relationship in the delivery of service. the success of the customer experience. Perhaps this will change with time as more firms start experimenting with their service delivery design as 2.1. Aspects of Service Design Bank of America has been doing (Thomke 2003). Service research has usually focused on capacity man- agement (type of customer contact, scheduling, and 2.3. Descriptive vs. Prescriptive Studies of Financial deployment) and the impact of the response to vari- Services ability on costs and quality. For long the nature of Several descriptive studies have focused on retail customer contact has influenced service design think- banking (Menor and Roth 2008, Menor et al. 2001), Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 639 substitution of labor with information technology environmental security, operational competence, and (Fung 2008), the use of customer feedback to improve quality of service help create trust. customer satisfaction (Krishnan et al. 1999), the use In general, service quality is difficult to manage and of distribution channels (Lee et al. 2004, Xue et al. measure because of the variability in customer expec- 2007), self-service technologies (such as ATMs, tations, their involvement in the delivery of the pay at the pump, see Campbell and Frei 2010a, b, service, etc. In general, there may be two different Meuter et al. 2000), online banking (Hitt and measures of service quality that are commonly used: Frei 2002), and e-services in general (see Boyer et al. the first refers to and measures the actual service pro- 2002, Ciciretti et al. 2009, Clemons et al. 2002, Furst vided (e.g., customer satisfaction, resolution, etc.), the et al. 2002, Menor et al. 2001). These studies talk about second may refer to the availability of service capac- the types of customers who use the various different ity/personnel (e.g., service level, availability, waiting channels and how firms have diversified their deliv- time, etc.). The first type of quality measure is not as ery of services using these new channels as newer nebulous in financial services where the output is technologies have become available. However, they generally related to monetary outcomes. If there is an are usually descriptive, rather than prescriptive, in error in the posting of a transaction, or if quarterly that they speak about how existing firms and cus- returns from a mutual fund are below industry per- tomers have already adopted these technologies, formance, there is an immediate customer reaction rather than what they should be doing in the future. and the points in the service design that caused such For example, there are few quantitative metrics to failures to occur is apparent, whether it is in remit- measure a product (e.g., its complexity vis-a-vis tance processing or in the hiring of a fund manager. customer knowledge), a process (e.g., face to face vs. Quality in financial services is not influenced by such automated), and proximity (on-site or off-site) to help matters as the mood of the customer, as may be the a manager navigate financial service operations strat- case in other services. This makes ensuring quality in egies from a design standpoint based on where her financial services more doable and one of the foci of firm is now. In that sense, financial service system the research in operational risk management which design still has ways to go to catch up with product we will discuss later. design (product attributes, customer utility, pricing, Roth and Jackson (1995) found that market intelli- form and function, configuration, product develop- gence and imitation of best practices can be an ment teams, etc.) and manufacturing process design effective way of improving service quality, and that (process selection, batch/line, capacity planning, service quality is more influenced by service process rigid/flexible automation, scheduling, location analy- choices and the cumulative impact of investments sis, etc.). Because batching and lot sizing issues have than by people’s capabilities. Productivity measure- been of considerable interest in the history of the ment in services is also a challenge (Sampson and study of manufacturing processes, and because online Froehle 2006). Bank performance as a result of process technologies have made the concept of batching con- variation has been studied by Frei et al. (1999). siderably less important, it would be interesting to see This current special issue of Production and Opera- how research in service systems design unfolds in the tions Management provides some interesting new future. One paper with prescriptive recommendations cases of process improvement in financial services. for service design in the property casualty insurance The paper by Apte et al. (2010), ‘‘Analysis and industry is due to Giloni et al. (2003). improvement of information-intensive services: Evi- dence from insurance claims handling operations,’’ 3. Financial Services Performance presents a classification of information-intensive Measurement and Analysis services based on their operational characteristics; this paper proposes an empirically grounded concep- 3.1. Best Practices and Process Improvement tual analysis and prescriptive frameworks that can be Many service firms are measuring success by factors used to improve the performance of information- and other than profitability, using such factors as customer customer contact-intensive services. The paper by De and employee loyalty, as measured by retention, Almeida Filho et al. (2010) focuses on collection pro- depth of relationship, and lifetime value (Heskett cesses in consumer credit. They develop a dynamic et al. 1994). Chen and Hitt (2002), in an empirical programming model to optimize the collections pro- study on retention in the online brokerage industry, cess in consumer credit. Collection processes have found that ease of use, breadth of offerings, and qual- been the Cinderella of consumer lending research, ity reduce customer attrition. Balasubramanian et al. because psychologically lenders do not enjoy analyz- (2003) find that trust is important for online transac- ing their mistakes, and also once an accounting loss is tions, because physical appearance of branches, etc. ascribed to a defaulted loan, there had been little no longer matter in such situations. Instead, perceived incentive for senior managers to keep track of how Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 640 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

Figure 1 Typical Structure of an Asset Management Organization

Asset management Independent internal oversight o Investment research, management, functions and execution o Compliance, legal and regulatory o Sales and client relationship o Controllers management o Credit and market risk o Product development management o Marketing o Internal audit o Valuation oversight

Internal support teams External service providers o Billing o Brokerage, clearing and execution o Human resources o Custody and trust services o Operations o Fund administrator o Operational risk o Prime brokerage and financing o Performance o Reputable auditor o Tax o Valuation (reputable third-party o Technology valuation firm) o Treasury

much will be subsequently collected. The paper by the multiple constituent parts that must work together Buell et al. (2010) investigates why self-service in order for a typical asset management organization customers are more reluctant to change their service to function effectively. Figure 2, also adapted from provider. This paper’s primary contribution is to Alptuna et al. (2010), lists the functions in the invest- investigate how satisfaction and switching costs con- ment management process according to their distance tribute to retention among self-service customers. This from the end client. Typically, the operations-intensive is a particularly important issue in the financial ser- functions reside in the middle and back offices; vices industry where considerable investments have accordingly, the untapped research potential of oper- been made in developing self-service distribution ations in asset management must be sought there. One channels, and migrating customers to them. can create a similar framework, as shown in Figures 1 and 2, for a typical retail bank, credit card issuer, 3.2. An Example of Best Practices: Asset mortgage lender, brokerage, trust bank, asset custo- Management dian, life or property/casualty insurer, among others, Asset management provides an interesting example of none of which is less complex than an asset manager. an area within the financial services sector that has Outsourcing operations adds to the complexity by in- been receiving an increasing amount of research troducing elements of quality control for outsourced attention with regard to best practices from various pieces and coordination between the main organiza- operations management perspectives. The body of tion and the third-party provider (State Street 2009). research on operations management in asset manage- To develop their framework, Alptuna et al. (2010) ment is growing, however, not always produced by draw heavily on asset management industry resources operations management researchers, but often by on best practices, namely the Managed Funds Associ- those in the finance world (Black 2007, Brown et al. ation’s Sound Practices for Hedge Fund Managers 2009a, b, Kundro and Feffer 2003, 2004, Stulz 2007), (2009), the Report of the Asset Managers’ Committee to who examine operational risk issues in hedge funds. the President’s Working Group on Financial Markets A collection of operations management research (2009), the Alternative Investment Management Asso- papers in asset management can be found in a recent ciation’s Guide to Sound Practices for European Hedge book by Pinedo (2010). Alptuna et al. (2010) present a Fund Managers (2007), and the CFA Institute’s Asset best practices framework for the operational infra- Manager Code of Professional Conduct (2009). structure and controls in asset management and argue Schneider (2010) provides a framework for asset that it is possible to effectively implement such a management firms to analyze their costs. Arfelt (2010) framework in organizations that enjoy a strong, prin- proposes an adaptation of the Lean Six Sigma frame- ciple-based governance. They examine conditions work used in automobile manufacturing for asset under which the cost-effective strategy of outsourc- management. Biggs (2010) advocates a decentraliza- ing asset management operations can be successful tion of risk management, accountability as well as for asset managers and their clients. Figure 1, which technology and expense control in asset management has been adapted from Alptuna et al. (2010), shows firms. Cruz (2010) argues that the focus of cost man- Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 641

Figure 2 Investment Management Process Functions

Front Asset management Trade execution office - Investment research - Financial - Portfolio and risk Information management eXchange (FIX) - Sales and client connectivity relationship - Trade order management management and - Product development execution Middle Investment operations office - Billing - Corporate actions - Performance - Cash administration processing and analytics - Client data warehouse - Data management - Portfolio - Client reporting - OTC derivatives recordkeeping processing and accounting - Reconciliation processing - Transaction management Back office Fund accounting Global custody Transfer agency - Daily, monthly, and ad- - Assets safekeeping - Shareholder hoc reporting - Cash availability servicing - General ledger - Failed trade - NAV calculation reporting - Reconciliation - Income/tax - Security pricing reclaims - Reconciliation - Trade settlement agement programs at asset management firms should Zenios et al. (1999), Soteriou and Zenios (1999b), and be strategic and tactical (see also Cruz and Pinedo Athanassopoulos and Giokas (2000) for Greek banks; 2009). Nordgard and Falkenberg (2010) give an IT Kantor and Maital (1999) for a large Mideast bank; perspective on costs in asset management. Campbell and Berger and Humphrey (1997) for various inter- and Frei (2010a) examine cost structure patterns national financial services firms. These papers discuss in the asset management industry. Amihud and operational efficiency, profitability, quality, stock mar- Mendelson (2010) examine the effect of transaction ket performance, and the development of better cost costs on asset management, and study their implica- estimates for banking products via DEA. Cummins tions for portfolio construction, fund design, trade et al. (1999) use DEA to explore the impact of implementation, cash and liquidity management, and organizational form on firm performance. They com- customer acquisition and development strategies. pare mutual and stock property liability companies and find that in using managerial discretion and cost- 3.3. Performance Analysis Through Data efficiency stock companies perform better, and in lines Envelopment Analysis (DEA) of insurance with long payouts mutual companies There are numerous studies on performance and pro- perform better. ductivity analyses of retail banking that are based on Cook and Seiford (2009) present an excellent over- DEA. DEA is a technique for evaluating productivity view of the DEA developments over the past 30 years measures that can be applied to service industries in and Cooper et al. (2007) provide a comprehensive general. It compares productivity measures of differ- textbook on the subject. For a good survey and ent entities (e.g., bank branches) within the same cautionary notes on the pitfalls of improper interpre- service organization (e.g., a large retail bank) to one tation and use of DEA results (e.g., loosely using the another. Such a comparative analysis then boils down results for evaluative purposes when uncontrollable to the formulation of a fractional linear program. DEA variables exist), see Metters et al. (1999). Zhu (2003) has been used in many retail banks to compare discusses methods to solve imprecise DEA (IDEA), productivity measures of the various branches with where data on inputs and outputs are either bounded, one another. Sherman and Gold (1985), Sherman and ordinal, or ratio bounded, where the original linear Ladino (1995), and Seiford and Zhu (1999) performed programming DEA formulation can no longer be used. such studies for US banks; Oral and Yolalan (1990) Koetter (2006) discusses the stochastic frontier performed such a study for a bank in Turkey; Vassi- analysis (SFA) as another bank efficiency analysis loglou and Giokas (1990), Soteriou and Zenios (1999a), framework, which contrasts to the deterministic DEA. Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 642 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

4. Forecasting ops models for different deposit types, i.e., business and Forecasting is very important in many areas of the personalchecking,NOW,savings,andmoneymarket financial services industry. In its most familiar form in account deposits. which it presents itself to customers and the general Labe and Papadakis (2010) discuss a propensity public, it consists of economic and market forecasts score matching model that can be used to forecast the developed by research and strategy groups in broker- likelihood of Bank of America’s retail clients bringing in age and investment management firms. However, the new funds to the firm by subscribing to promotional types of forecasting we discuss tend to be more inter- offerings of CDs. Such promotional CDs carry an nal to the firms and not visible from the outside. above-market premium rate for a limited period of time. Humphrey et al. (2000) forecast the adoption of 4.1. Forecasting in the Management of Cash electronic payments in the United States; they find that Deposits and Credit Lines one of the reasons for the slow pace of moving from Deposit-taking institutions (e.g., commercial banks, checks to electronic payments in the United States is the savings and loan associations, and credit unions) are customers’ perceived loss of float. Many electronic pay- interested in forecasting the future growth of their ment systems now address this, by allowing for deposits. They use this information in the process of payment at the due date rather than immediately. determining the value and pricing of their deposit Revolving credit lines, or facilities, give borrowers products (e.g., checking, savings, and money market access to cash on demand for short-term funding needs accounts, and also CDs), for asset–liability manage- up to credit limits established at facility inception. Banks ment, and for capacity considerations. Of special typically offer these facilities to corporations with in- interest to these institutions are demand deposits, vestment grade credit ratings, which have access to more broadly defined as non-maturity deposits. cheaper sources of short-term funding, for example, Demand deposits have no stated maturity and the commercial paper, and do not draw significant amounts depositor can add to the balance without restriction, from them except: or withdraw from ‘‘on demand,’’ i.e., without warning (i) for very brief periods of time under normal or penalty. In contrast, time deposits, also known as conditions, CDs, have a set maturity and an amount established (ii) when severe deterioration of their financial at inception, with penalties for early withdrawals. condition causes them to lose access to the Forecasting techniques have been applied to demand credit markets, and deposits because of their relative non-stickiness due to (iii) during system-wide credit market dysfunction, the absence of contractual penalties. A product with such as during the crisis of 2007–2009. similar non-stickiness is credit card loans. Jarrow and Van Deventer’s (1998) model for valuing demand Banks that offer these credit facilities must set aside deposits and credit card loans using an arbitrage-free adequate, but not excessive, funds to satisfy the de- methodology assumes that demand deposit balances mand for cash by facility borrowers. Duffy et al. (2005) depend only on the future evolution of interest rates; describe a Monte Carlo simulation model that Merrill however, it does allow for more complexity, such as Lynch Bank used to forecast these demands for cash macroeconomic variables (income or unemployment), by borrowers of their revolver portfolio. The model and local market or firm-specific idiosyncratic factors. uses industry data for revolver usage by borrower Janosi et al. (1999) use a commercial bank’s demand credit rating, and assumes Markovian credit rating deposit data and aggregate data for negotiable order migrations, correlated within and across industries. of withdrawal (NOW) accounts from the Federal Migration probabilities were provided by a major Reserve to empirically investigate Jarrow and Van Dev- rating agency, and correlation estimates were calcu- enter’s model. They find demand deposit balances to be lated by Merrill Lynch’s risk group. The model was strongly autoregressive, i.e., future balances are highly used by Merrill Lynch Bank to help manage liqui- correlated with past balances. They develop regression dity risk in its multi-billion portfolio of revolving models, linear in the logarithm of balances, in which credit lines. past balances, interest rates, and a time trend are pre- Forecasting the future behavior and profitability of dictive variables. O’Brien (2000) adds income to the set retail borrowers (e.g., for credit card loans, mortgages, of predictive variables in the regression models. Shee- and home equity lines of credit) has become a key han (2004) adds month-of-the-year dummy variables in component of the credit management process. Fore- the regressions to account for calendar-specific inflows casting involved in a decision to grant credit to a new (e.g., bonuses or tax refunds), or outflows (e.g., tax borrower is known as ‘‘credit scoring,’’ and its origins payments). He focuses on core deposits, i.e., checking in the modern era can be found in the 1950s. A dis- accounts and savings accounts; distinguishes between cussion of credit scoring models, including related the behavior of total and retained deposits; and devel- public policy issues, is offered by Capon (1982). Fore- Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 643 casting involved in the decisions to adjust credit ac- retail brokerage to forecast client choice decisions on cess and marketing effort to existing borrowers is introduction of lower-cost offerings to complement known as ‘‘behavioral scoring.’’ The book by Thomas the firm’s traditional full-service channel. Client et al. (2002) contains a comprehensive review of the choice decision forecasts were used as inputs in the objectives, methods, and practical implementation of process of determining the proper pricing for these credit and behavioral scoring. The formal statistical new offerings and for evaluating their potential methods used for classifying credit applicants into impact on firm revenue. The results of a rational eco- ‘‘good’’ and ‘‘bad’’ risk classes is known as ‘‘classifi- nomic behavior (REB) model were used as a baseline. cation scoring.’’ Hand and Henley (1997) review The REB model assumes that investors optimize their a significant part of the large body of literature in value received by always choosing the lowest-cost classification scoring. Baesens et al. (2003) examine option (determined by an embedded optimization the performance of standard classification algorithms, model that was solved for each of millions of clients including logistic regression, discriminant analysis, and their actual portfolio holdings). The REB model’s k-nearest neighbor, neural networks, and decision trees; results were compared with those of a Monte Carlo they also review more recently proposed ones, such as simulation model. The Monte Carlo simulation allows support vector machines and least-squares support for more realistic assumptions. For example, clients’ vector machines (LS-SVM). They find LS-SVM, and decisions are impacted not only by price differentials neural network classifiers, and simpler methods such as across channels, but also by the strength and quality logistic regression and linear discriminant analysis to of the relationship with their financial advisor, who have good predictive power. In addition to classifica- represented the higher-cost options. tion scoring, other methods include: Labe (1994) describes an application of forecasting the likelihood of affluent prospects becoming Merrill (i) ‘‘response scoring,’’ which aims to forecast a Lynch’s priority brokerage and investment advisory prospect’s likelihood to respond to an offer for clients (defined as clients with more than US$250,000 credit, and in assets). Merrill Lynch used discriminant analysis, a (ii) ‘‘balance scoring,’’ which forecasts the pros- method akin to classification scoring, to select high pect’s likelihood of carrying a balance if they quality households to target in its prospecting efforts. respond. The trading of securities in capital markets involves To improve the chances of acquiring and maintaining key operational functions that include: profitable customers, offers for credit should be mailed (i) clearing, i.e., establishing mutual obligations of only to prospects with high credit, response, and bal- counterparties in securities and/or cash trades, as ance scores. Response and balance scoring models are well as guarantees of payments and deliveries, typically proprietary. Trench et al. (2003) discuss a and model for optimally managing the size and pricing of (ii) settlement, i.e., transfer of titles and/or cash to card lines of credit at Bank One. The model uses the accounts of counterparties in order to final- account-level historical transaction information to select ize transactions. for each cardholder, through Markov decision processes, annual percentage rates, and credit lines that optimize Most major markets have centralized clearing the net present value of the bank’s credit portfolio. facilities so that counterparties do not have to settle bilaterally and assume credit risk to each other. The 4.2. Forecasting in Securities Brokerage, Clearing, central clearing organization must have robust pro- and Execution cedures to satisfy obligations to counterparties, i.e., In the last few decades, the securities brokerage minimize the number of trades for which delivery of industry has seen dramatic change. Traditional wire- securities is missed. It must also hold adequate, but houses charging fixed commissions evolved or were not excessive, amounts of cash to meet payments. replaced by diverse organizations offering full service, Forecasting the number and value of trades during a discount, and online trading channels, as well as re- clearing and settlement cycle can help the organiza- search and investment advisory services. This tion meet the above objectives; it can achieve this by evolution has introduced a variety of channel choices modeling the clearing and settlement operation using for retail and institutional investors. Pricing, service stochastic simulation. A different approach is used by mix and quality, and human relationships are key de Lascurain et al. (2011): they develop a linear pro- determinants in the channel choice decision. Firms are gramming method to model the clearing and settlement interested in forecasting channel choice decisions by operation of the Central Securities Depository of clients, because they greatly impact capacity planning, Mexico and evaluate the system’s performance through revenue, and profitability. Altschuler et al. (2002) dis- deterministic simulation. The model’s formulation in de cuss simulation models developed for Merrill Lynch’s Lascurain et al. (2011) is a relaxation of a mixed integer Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 644 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society programming (MIP) formulation proposed by Gu¨ntzer parison to the variability expected in a Poisson et al. (1998), who show that the bank clearing problem process and can have significant impact on the pre- is NP-complete. Eisenberg and Noe (2001) include dictions of long-run performance; ignoring forecast clearing and settlement in a systemic financial risk errors typically leads to overestimation of perfor- framework. mance. Jongbloed and Koole (2001) found that the call arrival data they had been analyzing had a variance 4.3. Forecasting of Call Arrivals at Call Centers much greater than the mean, and therefore did not Forecasting techniques are also used in various other appear to be samples of Poisson distributed random areas within the financial services sector; for example, variables. They addressed this ‘‘overdispersion’’ by the forecasting of arrivals in call centers, which is a proposing a Poisson mixture model, i.e., a Poisson crucial input in the personnel scheduling process model with an arrival rate that is not fixed but ran- in call centers (to be discussed in a later section). To dom following a certain stochastic process. Brown set this in a broader context, refer to the framework of et al. (2005) found data from a different call center that Thompson (1998) for forecasting demand for services. also followed a Poisson distribution with a variable Recent papers that focus on forecasting call center arrival rate; the arrival rates were also serially corre- workload include a tutorial by Gans et al. (2003), a lated from day to day. The prediction model proposed survey by Aksin et al. (2007), and a research paper by includes the previous day’s call volume as an auto- Aldor-Noiman et al. (2009). regressive term. High intra-day correlations were The quality of historical data improves with the found by Avramidis et al. (2004), who developed passage of time because call centers become increas- models in which the call arrival rate is a random ingly more sophisticated in capturing data with every variable correlated across time intervals of the same nuance that a modeler may find useful or interesting. day. Steckley et al. (2004) and Mehrotra et al. (2010) Andrews and Cunningham (1995) describe the auto- examine the correlation of call volumes at later peri- regressive integrated moving average (ARIMA) fore- ods of a day to call volumes experienced earlier in the casting models used at L.L. Bean’s call centers. Time day for the purpose of updating workload schedules. series data used to fit the models exhibit seasonality Methods to approximate non-homogeneous Pois- patterns and are also influenced by variables such as son processes often attempt to estimate the arrival rate holiday and advertising interventions. Advertising by breaking up the data set into smaller intervals. and special calendar effects are addressed by Antipov Henderson (2003) demonstrates how a heuristic that and Meade (2002). More recently, Soyer and assumes a piecewise constant arrival rate over time Tarimcilar (2008) incorporate advertising effects by intervals with a length that shrinks as the volume modeling call arrivals as a modulated Poisson process of data grows produces good arrival rate function with arrival rates being driven by customer calls that estimates. Massey et al. (1996) fit piecewise linear rate are stimulated by advertising campaigns. They use a functions to approximate a general time-inhomoge- Bayesian modeling framework and a data set from a neous Poisson process. Weinberg et al. (2007) forecast call center that enables tracing calls back to specific an inhomogeneous Poisson process using a Bayesian advertisements. In a study of Fedex’s call centers, framework, whereby from a set of prior distributions Xu (2000) presents forecasting methodologies used they estimate the parameters of the posterior dis- at multiple levels of the business decision-making tribution through a Monte Carlo Markov Chain hierarchy, i.e., strategic, business plan, tactical, and model. They forecast arrival rates in short intervals operational, and discusses the issues that each meth- of 15–60 minutes of a day of the week as the product odology addresses. Methods used include exponential of a day’s forecast volume times the proportion of smoothing, ARIMA models, linear regression, and calls arriving during an interval; they also allow for a time series decomposition. random error term. At low granularity, call arrival data may have too Shen and Huang (2005, 2008a, b) developed models much noise. Mandelbaum et al. (2001) demonstrate for inter-day forecasting and intra-day updating of how to remove relatively unpredictable short-term call center arrivals using singular value decomposi- variability from data and keep only predictable vari- tion. Their approach resulted in a significant dimen- ability. They achieve this by aggregating data at sionality reduction. In a recent empirical study, Taylor higher levels of granularity, i.e., progressively mov- (2008) compared the performance of several univari- ing up from minute of the hour to hour of the day ate time series methods for forecasting intra-day call to day of the month and to month of the year. The arrivals. Methods tested included seasonal autore- elegant textbook assumption that call arrivals follow a gressive and exponential smoothing models, and the Poisson process with a fixed rate that is known or can dynamic harmonic regression of Tych et al. (2002). be estimated does not hold in practice. Steckley et al. Results indicate that different methods perform best (2009) show that forecast errors can be large in com- under different lead times and call volumes levels. Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 645

Forecasting other aspects of a call center with a The concept of transactions demand for money, significant potential for future research include, for addressed by the Baumol–Tobin model, is related to, example, waiting times of calls in queues, see Whitt but subtly different from, precautionary demand for (1999a, b) and Armony and Maglaras (2004). cash that applies to unforeseen expenditures, oppor- tunities for advantageous purchases, and uncertainty 5. Inventory and Cash Management in receipts. Whalen (1966) developed a model with a structure strikingly similar to the Baumol–Tobin 5.1. Cash Inventory Management Under model, capturing the stochastic nature of precaution- Deterministic and Stochastic Demand ary demand for cash. Sprenkle (1969) and Akerlof and Organizations, households, and individuals need cash Milbourne (1978) observed that the Baumol–Tobin to meet their liquidity needs. In the era of checks and model tends to under-predict demand for money, electronic transactions, an amount of cash does not partly because it fails to capture the stochastic nature have to be in physical currency, but may correspond of precautionary demand for cash. Sprenkle’s paper only to a value in an account that has been set up for elicited a response by Orr (1974), which in turn this purpose. To meet short-term liquidity needs, cash prompted a counter-response by Sprenkle (1977). must be held in a riskless form, where its value does Robichek et al. (1965) propose a deterministic short- not fluctuate and is available on demand, but earns term financing model that incorporates a great degree little or no interest. Treasury bills and checking of realistic detail involved in the financial officer’s accounts are considered riskless. Cash not needed to decision-making process, which they formulate and meet short-term liquidity needs can be invested in solve as a linear program. They include a discussion risky assets whereby it may earn higher returns, but on model extensions for solving the financing prob- its value may be subject to significant fluctuations and lem under uncertainty. Sethi and Thompson (1970) uncertainty, and could become wholly or partially proposed models based on mathematical control the- unrecoverable. Depending on the type of risky asset, ory, in which demand for cash is deterministic but its value may or may not be quickly recoverable and does vary with time. In an extension of the Sethi– realizable at a modest cost (as with a public equity Thompson model, Bensoussan et al. (2009) allow the that is listed in a major stock exchange). Determining demand for cash to be satisfied by dividends and the value of certain types of risky assets (e.g., private uncertain capital gains of the risky asset, stock. equity, real estate, some hedge funds, and asset- In what became known as the Miller–Orr model backed fixed income securities) may require special- for cash management, Miller and Orr (1966) extended ized valuation services, which could involve the Baumol–Tobin model by assuming the demand for significant time and cost. Risky assets can also be cash to be stochastic. The cash balance can fluctuate subject to default, in which case all or part of the value randomly between a lower and an upper bound becomes permanently unrecoverable. according to a Bernoulli process, and transactions take Researchers have produced over the last few decades place when it starts moving out of this range; units of a significant body of work by applying the principles of the risky asset are converted into cash at the lower inventory theory to cash management. We review the bound, and bought with the excess cash at the upper cash management literature from its beginnings so bound. Transaction costs were assumed fixed, we can put later work in context, and we have not i.e., independent of transaction size. In a critique of identified an earlier comprehensive review that accom- the Miller–Orr model, Weitzman (1968) finds it to be plishes this purpose. Whistler (1967) discussed a ‘‘robust,’’ i.e., general results do not change much stochastic inventory model for rented equipment that when the underlying assumptions are modified. wasformulatedasadynamicprogram;thiswork Eppen and Fama (1968, 1969, 1971) proposed cash served as a model for the cash management problem. balance models that are embedded in a Markovian One of the early works produced an elegant result that framework. In one of their papers, Eppen and Fama became known as the Baumol–Tobin economic model (1969) presented a stochastic model, formulated as a of the transactions demand for money, independently dynamic program, with transaction costs proportional developed by Baumol (1952) and Tobin (1956). to transaction sizes. Changes in the cash balance can The model assumes a deterministic, constant rate of follow any discrete and bounded probability distribu- demand for cash. It calculates the optimal ‘‘lot sizes’’ of tion. In another one of their papers, Eppen and Fama the risky asset to be converted to cash, or the optimal (1968) developed a general stochastic model that numbers of such conversions, in the presence of trans- allowed costs to have a fixed as well as a variable com- action and interest costs. Tobin’s version requires an ponent. They showed how to find optimal policies for integer number of transactions and therefore approxi- the infinite-horizon problem using linear programming. mates reality more closely than Baumol’s, which allows In their third paper, Eppen and Fama (1971) proposed a that variable to be continuous. stochastic model with two risky assets, namely ‘‘bonds’’ Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 646 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society and ‘‘stock’’; the stock is more risky but has a higher and Fase (1992) present studies by the Dutch central expected return. They also discussed using ‘‘bonds’’ bank on the demand for banknotes in the Netherlands (the intermediate-risk asset) as a ‘‘buffer’’ between cash before the introduction of the Euro. Ladany (1997) and the more risky asset. Taking a similar approach, developed a discrete dynamic programming model to Daellenbach (1971) proposed a stochastic cash balance determine optimal (minimum cost) ordering policies model using two sources of short-term funds. Girgis for banknotes by Israel’s central bank. Massoud (2005) (1968) and Neave (1970) presented models with both presents a dynamic cost minimizing note inventory fixed and proportional costs and examined conditions model to determine optimal banknote order size and for policies to be optimal under different assumptions. frequency for a typical central bank. Hausman and Sanchez-Bell (1975) and Vickson (1985) The production and distribution of banknotes, and developed models for firms facing a compensating- the required infrastructure and processes, have also balance requirement specified as an average balance been studied. Fase et al. (1979) discuss a numerical over a number of days. planning model for the banknotes operations at a Continuous-time formulations of the cash manage- central bank, with examples from pre-Euro Nether- ment problem were based on the works of Antelman lands. Bauer et al. (2000) develop optimization models and Savage (1965), and Bather (1966), who used a for determining the least-cost configuration of the US Wiener process to generate a stochastic demand in Federal Reserve’s currency processing sites given the their inventory problem formulations. Their approach trade-off between economies of scale in processing was extended to cash management by Vial (1972), and transportation costs. In a study of costs and econ- whose continuous-time formulation had fixed and omies of scale of the US Federal Reserve’s currency proportional transaction costs, and linear holding and operations, Bohn et al. (2001) find that the Federal penalty costs, and determined the form of the optimal Reserve is not a natural monopoly. Opening currency policy (assuming one exists). Constantinides (1976) operations to market competition and charging fees extended the model by allowing positive and negative and penalties for some services provided for free by cash balances, determined the parameters of the the Federal Reserve at that time could lead to more optimal policy, and discussed properties of the efficient allocation of resources. optimal solution. Constantinides and Richard (1978) The movement of physical cash among central formulated a continuous-time, infinite-horizon, banks, depository institutions, and the public must be discounted-cost cash management model with fixed studied as a closed-loop supply chain (see, e.g., and proportional transaction costs, linear holding and Dekker et al. 2004). It involves the recirculation of penalty costs, and the Wiener process as the used notes back into the system (reverse logistics), demand-generating mechanism. They proved that there together with a flow of new notes from the central always exists an optimal policy for the cash manage- bank to the public through depository institutions ment problem, and that this policy is of a simple form. (forward logistics). The two movements are so inter- Smith (1989) developed a continuous-time model with twined that they cannot be decoupled. Rajamani et al. a stochastic, time-varying interest rate. (2006) study the cash supply chain structure in the United States, analyze it as a closed-loop supply 5.2. Supply Chain Management of Physical chain, and describe the cash flow management system Currency used by US banks. They also discuss the new cash Physical cash, i.e., paper currency and coins, remains recirculation policies adopted by the Federal Reserve an important component of the transactions volume to discourage banks’ overuse of its cash processing even in economies that have experienced a significant services, and encourage increased recirculation at the growth in checks, credit, debit and smart cards, and depository institution level. Among the practices to be electronic transactions. Advantages of cash include discouraged was ‘‘cross shipping,’’ i.e., shipping used ease of use, anonymity, and finality; it does not currency to the Federal Reserve and ordering it in the require a bank account; it protects privacy by leaving same denominations in the same week. To compare no transaction records; and it eliminates the need to and contrast new and old Federal Reserve policies for receive statements and pay bills. Disadvantages currency recirculation, Geismar et al. (2007) introduce of cash include ease of tax evasion, support of an models that explain the flow of currency between the ‘‘underground’’ economy, risk of loss through theft or Federal Reserve and banks under both sets of guide- damage, ability to counterfeit, and unsuitability for lines. They present a detailed analysis that provides online transactions. optimal policies for managing the flow of currency Central banks provide cash to depository institu- between banks and the Federal Reserve, and analyze tions, which in turn circulate it in the economy. There banks’ responses to the new guidelines to help the are studies on paper currency circulation in various Federal Reserve understand their implications. countries. For example, Fase (1981) and Boeschoten Dawande et al. (2010) examine the conditions that Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 647 can induce depository institutions to respond in so- account in order to cover potential future transactions cially optimal ways according to the new Federal and avoid the need for a sixth sweep. Reserve guidelines. Mehrotra et al. (2010), which is a The impact of the sequence of transaction postings paper in this special issue of Production and Operations on account balances and resultant fees for insufficient Management, address the problem of obtaining effi- funds, similar to the cost of stock-outs in inventory cient cash management operating policies for management, has been studied by Apte et al. (2004). depository institutions under the new Federal Re- They investigate how overdraft fees and non- serve guidelines. The mixed-integer programming sufficient funds (NSF) fees interact in such situations. model developed for this purpose seeks to find Brokerage houses make loans to investors who want ‘‘good’’ operating policies, if such exist, to quantify to use leverage, i.e., to invest funds in excess of their the monetary impact on a depository institution op- own capital in risky assets, and can pledge securities erating according to the new guidelines. Another that they own as collateral. In a simple application of objective was to analyze to what extent the new this practice, known as margin lending, the brokerage guidelines can discourage cross shipping and stimu- extends a margin loan to a client of up to the value late currency recirculation at the depository of equity securities held in the client’s portfolio. The institution level. Mehrotra et al. (2010a) study pricing client can use the loaned funds to buy more equity and logistics schemes for services such as fit-sorting securities. Calculating the minimum value required in and transportation that can be offered by third-party a client’s account for a margin loan can become com- providers as a result of the Federal Reserve’s new plex in accounts holding different types of securities policies. including equities, bonds, and derivatives, all with different margin requirements. The complexity in- 5.3. Other Cash Management Applications in creases even more with the presence of long and short Banking and Securities Brokerage positions and various derivative strategies practiced US banks are required to keep on reserve a minimum by clients. Rudd and Schroeder (1982) presented a percentage (currently 10%) of deposits in client trans- simple transportation model formulation for calculat- action accounts (demand deposits and other ing the minimum margin, which represented an checkable deposits) at the Federal Reserve. Until very improvement over the heuristics used in practice. A recently, banks had a strong incentive to keep funds significant body of subsequent work has been pub- on reserve at a minimum, because these funds were lished on this problem, especially by Timkovsky and earning no interest. Even after the 2006 Financial Ser- collaborators, which is more related to portfolio strat- vices Regulatory Relief Act became law, authorizing egies and hedging. We believe that the approach in the payment of interest on reserves held at the Federal paper by Fiterman and Timkovsky (2001), which is Reserve, banks prefer to have funds available for their based on 0–1 knapsack formulations, is methodolog- own use rather than have them locked up on reserve. ically the most relevant to mention in this overview. Money market deposit accounts (MMDA) with check- ing allow banks to reduce the amounts on reserve at 6. Waiting Line Management in Retail the Federal Reserve by keeping deposits in MMDA accounts and transferring to a companion checking Banks and in Call Centers account only the amounts needed for transactions. 6.1. Queueing Environments and Modeling Only up to six transfers (‘‘sweeps’’) per month are Assumptions allowed from an MMDA to a checking account, and a In financial services, in particular in retail banking, client’s number and amount of transactions in the retail brokerage, and retail asset management (pen- days remaining in a month is unknown. Therefore, the sion funds, etc.), queueing is a common phenomenon size and timing of the first five sweeps must be care- that has been analyzed thoroughly. Queueing occurs fully calculated to avoid a sixth sweep, which will in the branches of retail banks with the tellers being move the remaining MMDA balance into the checking the servers, at banks of ATM machines with the account. Banks have been using heuristic algorithms machines being the servers, and in call centers, where to plan the first five sweeps. This specialized inven- the operators and/or the automated voice response tory problem has been examined by Nair and units are the servers. These diverse queueing envi- Anderson (2008), who propose a stochastic dynamic ronments turn out to be fairly different from one programming model to optimize retail account another, in particular with regard to the following sweeps. The stochastic dynamic programming model characteristics: developed by Nair and Hatzakis (2010) introduces cushions added to the minimum sweep amounts. It (i) the information that is available to the cus- determines the optimal cushion sizes to ensure tomer and the information that is available to that sufficient funds are available in the transaction the service system, Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 648 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

(ii) the flexibility of the service system with regard In the early 1980s retail banks began to make ex- to adjustments in the number of servers tensive use of ATMs. The ATMs at a branch of a bank dependent on the demand, behave quite differently from the human tellers. In (iii) the order of magnitude of the number of servers. contrast to a teller environment, the number of ATMs at a branch is fixed and cannot be adjusted as a Even though in the academic literature the arrival function of customer demand. However, the teller processes in queueing systems are usually assumed to environment and the ATM environment do have be stationary (time-homogeneous) Poisson processes, some similarities. In both environments, a customer arrival processes in practice are more appropriately can observe the length of the queue and can, therefore, modeled as non-homogeneous Poisson processes. Over estimate the amount of time (s)he has to wait. In the last couple of decades, some research has been done neither the teller environment, nor the ATM environ- on queues that are subject to non-homogeneous Poisson ment, can the bank adopt a priority system that would inputs (see, e.g., Massey et al. 1996). The more theoret- ensure that more valuable customers have a shorter ical research in queueing has also focused on various wait. Kolesar (1984) did an early analysis of a branch aspects of customer behavior in queue, in particular with two ATM machines and collected service time abandonment, balking, and reneging. For example, data as well as arrival time data. However, it became Zohar et al. (2002) have modeled the adaptive behav- clear very quickly that a bank of ATMs is capable of ior of impatient customers and Whitt (2006) developed collecting some very specific data automatically (e.g., fluid models for many-server queues with abandon- customer service times and machine idle times), but ment. In all three queueing environments described cannot keep track of certain other data (e.g., queue above, the psychology of the customers in the queue lengths, customer waiting times). Larson (1990), there- also plays a major role. A significant amount of research fore, developed the so-called queue inference engine, has been done on this topic, see Larson (1987), Katz et al. which basically provides a procedure for estimating (1991), and Bitran et al. (2008). As it turns out, reducing the expected waiting times of customers, given the wait times may not always be the best approach in all service times recorded at the ATMs as well as the service encounters. For example, in restaurants and machine idle times. salons, longer service time may be perceived as better service. In many cases, customers do not like waiting, 6.3. Waiting Lines in Call Centers but when it comes their turn to be served, would like Since the late 1980s, banks have started to invest the service to take longer. In still others, for example, in heavily in call center technologies. All major retail grocery checkout lines, customers want a businesslike banks now operate large call centers on a 24/7 basis. pace for both waiting and service. The latter category, Call centers have therefore been the subject of exten- which we may call dispassionate services, are more com- sive research studies, see the survey papers by Pinedo mon in financial service situations, though the former, et al. (2000), Gans et al. (2003), and Aksin et al. (2007). which we may call hedonic services, are also present—for The queueing system in a call center is actually quite example, when a customer visits their mortgage broker different from the queueing systems in a teller envi- or insurance agent, they would not like to be rushed. In ronment or in an ATM environment; there are a the following subsections, we consider the various number of major differences. First, a customer now different queueing environments in more detail. has no direct information with regard to the queue length and cannot estimate his waiting time; he must rely entirely on the information the service system 6.2. Waiting Lines in Retail Bank Branches and at provides him. On the other hand, the service organi- ATMs zation in a call center has detailed knowledge con- The more traditional queues in financial services cerning the customers who are waiting in queue. The are those in bank branches feeding the tellers. Such institution knows of each customer his or her identity a queue is typically a single line with a number of and how valuable (s)he is to the bank. The bank can servers in parallel. There are clearly no priorities in now put the customers in separate virtual queues such a queue and the discipline is just first come first with different priority levels. This new capability has served. Such a queueing system is typically modeled made the application of priority queueing systems as an M/M/s system and is discussed in many stan- suddenly very important; well-known results in dard queueing texts. One important aspect of this queueing theory have now suddenly become more type of queueing in a branch is that management applicable, see Kleinrock (1976). Second, the call usually can adjust the number of available tellers centers are in another aspect quite different from the fairly easily as a function of customer demand and teller and the ATM environments. The number time of day. (This gives rise to many personnel sched- of servers in either a teller environment or an ATM uling issues that will be discussed in the next section.) environment may typically be, say, at most 20, Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 649 whereas the number of operators in a call center may has increased the need for outbound calling from typically be in the hundreds or even in the thousands. call centers. Outbound calling is quite different from In the analysis of call center queues, it is now possible inbound calling, because the scheduling of calls is to apply limit theorems with respect to the number of done by the call center rather than the call center operators, see Halfin and Whitt (1981) and Reed being at the mercy of its customers. This presents (2009). Third, the banks have detailed information unique challenges and opportunities. Bollapragada with regard to the skills of each one of its operators in and Nair (2010) focus in this environment on a call center (language skills, product knowledge, improvements in contact rates of appropriate parties. etc.). This enables the bank to apply skills-based-rout- ing to the calls that are coming in, see Gans and Zhou 7. Personnel Scheduling in Retail (2003) and Mehrotra et al. (2009). Banks and in Call Centers In call centers, typically, an enormous amount of statistical data is available that is collected automat- 7.1. Preliminaries and General Research Directions ically, see Mandelbaum et al. (2001) and Brown et al. An enormous amount of work has been done on (2005). The data that are collected automatically are workforce (shift) scheduling in manufacturing. How- much more extensive than the data that are collected ever, workforce scheduling in manufacturing is quite in an ATM environment. It includes the waiting times different from workforce scheduling in services of the customers, the queue length at each point in industries. The workforce scheduling process in man- time, the proportion of customers that experience no ufacturing has to adapt itself to inventory wait at all, and so on. considerations and is typically a fairly regular and Lately, many other aspects of queueing in call cen- stable process. In contrast to manufacturing indus- ters have become the subject of research. This special tries, workforce scheduling in the service industries issue of Production and Operations Management as well has to adapt itself to a fluctuating customer demand, as another recent special issue contain several such which in practice is often based on non-homogeneous papers. For example, O¨ rmeci and Aksin (2010) focus Poisson customer arrival processes. In practice, adapt- on the effects of cross selling on the management of ing the number of tellers or operators to the demand the queues in call centers. They focus on the opera- process can be done through an internal pool of flex- tional implications of cross selling in terms of capacity ible workers, or through a partnership with a labor usage and value generation. Chevalier and Van den supply agency (see Larson and Pinker 2000). Schrieck (2008), and Barth et al. (2010) consider hier- As the assignment of tellers and the hiring of archical call centers that consist of multiple levels operators depend so strongly on anticipated customer (stages) with a time-dependent overflow from one demand, a significant amount of research has focused level to the next. For example, at the first stage, the on probabilistic modeling of arrival processes, on sta- front office, the customers receive the basic services; tistical analyses of arrival processes, and on customer a fraction of the served customers requires more spe- demand forecasting in order to accomplish a proper cialized services that are provided by the back office. staffing. For probabilistic modeling of customer Van Dijk and Van der Sluis (2008) and Meester et al. arrival processes, see Stolletz (2003), Ridley et al. (2010) consider the service network configuration (2004), Avramidis et al. (2004), and Jongbloed and problem. Meester et al. (2010) analyze networks of Koole (2001). For statistical analyses of customer geographically dispersed call centers that vary in ser- arrival data, see Brown et al. (2005) and Robbins vice and revenue-generation capabilities, as well as in et al. (2006). For customer demand forecasting, see costs. Optimally configuring a service network in this Weinberg et al. (2007) and Shen and Huang (2008a, b). context requires managers to balance the competing From a research perspective, the personnel sched- considerations of costs (including applicable dis- uling problem has been tackled via a number of counts) and anticipated revenues. Given the large different approaches, namely, simulation, stochastic scale of call center operations in financial services modeling, optimization modeling, and artificial intel- firms, the service network configuration problem is ligence. The application areas considered included the important and economically significant. The approach scheduling of bank tellers as well as the scheduling of by Meester and colleagues integrates decision prob- the operators in call centers. Slepicka and Sporer lems involving call distribution, staffing, and (1981), Hammond and Mahesh (1995), and Mehrotra scheduling in a hierarchical manner (previously, these and Fama (2003) used simulation to schedule bank decision problems were addressed separately). tellers and call center operators. Thompson (1993) Even though most call center research has focused studied the impact of having multiple periods with on inbound call centers, a limited amount of research different demands on determining the employee re- has also been done on outbound call centers. Rising quirements in each segment of the schedule; see also delinquencies and the importance of telemarketing Chen and Henderson (2001). Green et al. (2007) and Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 650 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

Feldman et al. (2008) have addressed the problem from Zeevi (2005) developed a staffing method for large call a stochastic point of view and have developed staffing centers based on stochastic fluid models. rules based on queueing theory. So et al. (2003) use Nowadays another aspect of personnel manage- better staff scheduling and team reconfiguration to ment in call centers has become important, namely improve check processing in the Federal Reserve Bank. call routing. As the clients may have many different demands and the operators may have many different skill sets, call routing has gained a significant amount 7.2. Optimization Models for Call Center Staffing of interest, see Mehrotra et al. (2009) and Bhulai et al. Many researchers have considered this problem from (2008). Optimal call routing typically has to be con- an optimization point of view. An optimization model sidered in conjunction with the optimal cross training to determine the optimal staffing levels and call rout- of the operators, see Robbins et al. (2007). ing can take a more aggregate form or a more detailed form. Bassamboo et al. (2005) consider a model with m 7.3. Miscellaneous Issues Concerning Workforce customer classes and r agent pools. They develop a Scheduling in Financial Services linear program-based method to obtain optimal Very little research has been done on workforce staffing levels as well as call routing. Not surpris- scheduling in other segments of the financial services ingly, a significant amount of work has also been done industry. It is clear that workforce scheduling is also on the more detailed optimization of the various pos- important in trading departments or at trading desks sible shift structures (including even lunch breaks and (equity trading, foreign exchange trading, or currency coffee breaks). In large call centers, there are many trading). Such departments have to keep track of in- different types of shifts that can be scheduled. A shift formation on trades, such as the number of trades is characterized by the days of the week, the starting done per day for each trader, amounts per trade and times at the beginning of the day, the ending time at total, and trades’ impact with regard to risk limits of the end of the day, as well as the timings of the lunch each trader or the entire desk. Automated systems breaks and coffee breaks. Gawande (1996) (see also exist for recording such information, but operational Pinedo 2009) solved this staffing problem in two risk mitigation practices necessitate the involvement stages. In the first stage of the approach, the so-called of skilled personnel in populating and reviewing of solid-tour scheduling problem is solved. (A solid tour the data. In the case of a broker/dealer, relevant represents a shift without any breaks; a solid tour is pieces of information may include the commission only characterized by the starting time at the begin- charged for each trade, whether the firm acted in a ning of the day and by the ending time at the end of capacity of principal or agent, and trading venue (e.g., the day.) There are scores of different solid tours an exchange, a dark pool, or an electronic communi- available (each one with a different starting time and cation network). Institutional and retail brokerage ending time). The demand process (the non-homoge- and asset management firms need to maintain neous arrival process) usually can be forecast with a databases with account-level, client-level, and rela- reasonable amount of accuracy. The first stage of the tionship-level information. Typically such information problem is to find the number of personnel to hire in is taken from forms filled out by clients on hard copy each solid tour such that the total number of people or electronically, and/or from legal agreements signed available in each time interval fits the demand curve by the client and the firm. Skilled personnel must as accurately as possible. This problem leads to an perform the back-office tasks of reviewing documents, integer programming formulation of a special struc- and populating the appropriate fields through user ture that actually can be solved in polynomial time. interfaces of databases. Given the regulatory implica- After it has been determined in the first stage what the tions of errors in this process (e.g., Sarbanes-Oxley 404 actual number of people in each solid tour is, the compliance), the personnel assigned to such tasks placement of the breaks is done in the second stage. must have received a rigorous specialized training. A Typically, the break placement problem is a very hard factor that may add complexity to workforce sched- problem, and is usually dealt with through a heuristic. uling is the emerging trend of outsourcing such Gans et al. (2009) applied parametric stochastic back-office tasks, especially to offshore locations in programming to workforce scheduling in call centers. Asia or Europe. In offshore outsourcing, time-zone Gurvich et al. (2010) have developed a chance con- differences and high turnover of trained professionals strained optimization approach to deal with uncertain can be serious issues to contend with. demand forecasts. Brazier et al. (1999) applied an As stated earlier, workforce scheduling and waiting artificial intelligence technique, namely co-operative line management are strongly interconnected. A multi-agent scheduling, on the workforce scheduling larger and better trained workforce clearly results in in call centers; their work was done in collaboration a better queueing performance. However, workforce with the Rabobank in the Netherlands. Harrison and scheduling is also strongly connected to operational Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 651 risk. A larger and better trained workforce clearly service industries and, in particular, process mapping results in a lower level of operational risk (especially in financial services. Process mapping also includes in trading departments, where human errors can have an identification of all potential failure points. This a very significant impact on the financial performance area of study is closely related to the research area of of the institution). This topic will be discussed in more reliability theory. detail in the next section on operational risk. The area of reliability theory is very much con- cerned with system and process design issues, including concepts such as optimal redundancies. 8. Operational Risk Management This area of research originated in the aviation indus- 8.1. Types of Operational Failures try; see the classic work by Barlow and Proschan Operational risk in financial services started to receive (1975) on reliability theory. One major issue in finan- attention from the banking community as well as from cial services is the issue of determining the amount of the academic community in the mid-1990s. Operational backups and the amount of parallel processing and risk has since then typically been defined as the risk checking that have to be designed into the processes resulting from inadequate or failed internal processes, and procedures. Reliability theory has always had a people, and systems, or from external events (Basel major impact on process design. Committee 2003). It covers product life cycle and exe- Procedures that are common in TQM turn out to be cution, product performance, information management very useful for the management of operational risk, and data security, business disruption, human re- for example, Six Sigma (see Cruz and Pinedo 2008). sources and reputation (see, e.g., the General Electric However, it has become clear that there are important Annual Report 2009, available at www.ge.com). Failures similarities as well as differences between TQM in the concerning internal processes can be due to transactions manufacturing industries and operational risk man- errors (some due to product design), or inadequate agement in financial services. One important difference operational controls (lack of oversight). Information sys- is that TQM in manufacturing (typically referred to as tem failures can be caused by programming errors or Six Sigma) is based on statistical properties of the Nor- computer crashes. Failures concerning people may be mal (Gaussian distribution), because any parameter due to incompetence or inadequately trained employ- that is being measured may have an equal probability ees, or due to fraud. Actually, since the mid-1990s a of having a deviation in one direction as in the other. number of events have taken place in financial services However, in operational risk management the analysis that can be classified as rogue trading. This type of of operational risk events focuses mainly on the out- event has turned out to be quite serious because just one liers, i.e., the catastrophic events. For that reason, such an event can bring down a financial services firm, the distributions used are different from the Normal; see Cruz (2002), Elsinger et al. (2006), Chernobai et al. they may be, for example, the Lognormal, or the Fat (2007), Cheng et al. (2007), and Jarrow (2008). Tail Lognormal. The statistical analysis often focuses on the occur- rences of rare, catastrophic events, because financial 8.2. Relationships between Operational Risk and services in general are quite concerned about being hit Other Research Areas by such rare catastrophic events. It is for this reason It has become clear over the last decade that the man- that EVT has become such an important research di- agement of operational risk in finance is very closely rection, see De Haan and Ferreira (2006). EVT is based related to other research areas in operations manage- on a limit theorem that is due to Gnedenko; this limit ment, operations research, and statistics. Most of the theorem specifies the distribution of the maximum of methodological research in operational risk has fo- a series of independent and identically distributed cused on probabilistic as well as statistical aspects of (i.i.d.) random variables. These extreme value distri- operational risk, for example, extreme value theory butions include the Weibull, Frechet, and Gumbel (EVT) (see Chernobai et al. 2007). The areas of impor- distributions. They typically have three parameters, tance include the following: namely one that specifies the mean, one that specifies (i) Process design and process mapping, the variance, and a third one that specifies the fatness (ii) Reliability theory, of the tail. The need for being able to parameterize the (iii) TQM, and fatness of the tail is based on the fact that the tail affects (iv) EVT. the probabilities of catastrophic events occurring.

The area of process design and process mapping is very important for the management of operational 8.3. Operational Risk in Specific Financial Sectors risk. As discussed in earlier sections of this paper, There are several sectors of the financial services Shostack (1984) focuses on process mapping in the industry that have received a significant amount of Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 652 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society attention with regard to their exposure to operational hedge funds that can be calculated from data in hedge risk, namely, fund databases. The purpose of the o score is to iden- tify problematic funds in a manner similar to Alt- (i) Retail financial services (banking, brokerage, man’s z score, which predicts corporate bankruptcies, credit cards), and can be used as a supplement for qualitative due (ii) Trading (equities, bonds, foreign exchange), diligence on hedge funds. In a subsequent study the (iii) Asset management (retail, institutional). same authors, Brown et al. (2009b), examine a com- There are several types of operational risk events that prehensive sample of due diligence reports on hedge are common in retail financial services. They include funds and find that misrepresentation, as well as not security breaches, fraud, and systems breakdowns. using a major auditing firm and third-party valuation, Security breaches in retail financial services often are key components of operational risk and leading involve clients’ personal information such as bank ac- indicators of future fund failure. count numbers and social security numbers. Such events can seriously damage an institution’s reputa- tion, bringing about punitive regulatory sanctions, and, 8.4. Other Research Directions in some well-known instances, becoming the basis for There are a number of other important research di- class-action lawsuits. There has been an extensive body rections that already have received some research of research in the literature that has focused on under- attention and that deserve more attention in the fu- standing the risk of information security (see Bagchi ture. First, how can we analyze the trade-offs between and Udo 2003, Dhillon 2004, Hong et al. 2003, Straub operational costs (productivity) and operational risk? and Welke 1998, Whitman 2004). Garg et al. (2003) have In the manufacturing and in the services literature, made an attempt to quantify the financial impact of some articles have appeared that discuss the trade- information security breaches. Fraud as a cause of an offs between costs and productivity on the one hand operational risk event is a major issue in the credit card and quality on the other hand; see, for example, Jones business. An enormous amount of work has been done (1988) and Kekre et al. (2009). However, this issue has on credit card fraud detection, mainly by researchers in not received much attention yet as far as the financial artificial intelligence and data mining; see, for example, services industry is concerned. Second, a fair amount Chan et al. (1999). Upgrades and consolidations of in- of research has been done with regard to mitigation formation systems may be major causes of operational of operational risk. The financial services industry risk events associated with systems breakdowns. has thoroughly studied what the aviation industry has Operational risk events in trading may be due to been doing with regard to operational risk, see Cruz human errors, rogue trading (i.e., either unauthorized and Pinedo (2008). In particular they have considered trades or illegal trades), or system breakdowns. A near-miss management practices of operational risk, number of such events have occurred in the last two see Muermann and Oktem (2002). Two more recent decades with catastrophic consequences for the insti- approaches for mitigating operational risk include in- tutions involved. Several rogue traders have caused surance (in particular with regard to rogue traders, their institutions billions of dollars in losses, see see Jarrow et al. 2010) and securitization (e.g., catas- Netter and Poulsen (2003). trophe bonds), see Cruz (2002). However, these issues Asset management has also been one of the areas concerning mitigation require more study. Third, it within the financial services sector that have recently has been observed that there is a significant interplay received a significant amount of research attention as between operational risk, market risk, and credit risk. far as operational risk is concerned. During the finan- For example, when the markets are very volatile, it cial crisis of 2007–2009, some of the most catastrophic is more likely that human errors may be made or losses suffered by investors resulted from failures to systems may crash. These correlations have to be properly address operational risk, for example, in the analyzed in more detail in the future. fraud perpetrated by Bernard Madoff (see Arvedlund The growing body of research on operational risk in 2009). Operational risk can be effectively addressed financial services presents interesting cross-fertilization by implementing robust operational infrastructures opportunities for operations management researchers, and controls in organizations that enjoy strong given the increasing visibility of operational risk and governance, as presented by Alptuna et al. (2010). the potential losses in financial services. For banks, the Operational risk is most salient in the loosely regu- Basel Capital Accord in a recent revision began to lated domain of hedge funds. Several studies have require risk capital to be reserved for potential shown that many of the hedge funds that have gone loss resulting from operational risk. Wei (2006) devel- under had major identifiable operational issues (e.g., oped models based on Bayesian credibility theory Kundro and Feffer 2003, 2004). Brown et al. (2009a) to quantify operational risk for firm-specific capital propose a quantitative operational risk score o for adequacy calculations. Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 653

9. Pricing and Revenue Management orientation has been published. In the rest of this sec- tion, we discuss the economic foundations of price 9.1. Pricing of Financial Services: Background and formation and attempt to link them to research in Academic Research finance on pricing practices specific to insurance, Financial services organizations expend serious efforts credit, and hedge funds. We also review the meager and resources on pricing and revenue management. operations management literature on pricing in finan- Applications are diverse; they include the setting of: cial services, examine the challenges faced by researchers, and propose links to other research that (i) interest rates (APR) on deposits and credit might help remedy some of the issues. products, (ii) trading commissions, (iii) custody fees, 9.2. Theory of Incentives and Informational Issues (iv) investment advisory fees, in Pricing of Financial Services (v) fund fees (which for hedge funds can be a By the neoclassical economic assumption of rational function of assets and performance), and individual behavior in perfect market competition, (vi) insurance policy premia. prices for financial products and services should be formed by: Pricing and revenue management are intertwined with many operations management functions in large (i) firms’ efforts to maximize profit, i.e., revenue financial services firms, because pricing strongly minus cost, and affects consumer demand for products and services, (ii) consumers’ desire to maximize their utility and customer attrition. Complicated pricing mecha- when faced with exogenous prices. nisms can increase the volume of billing questions to call centers. All of these can have significant implica- In this elegant model, information is perfectly known tions on how these products and services are best and shared by all economic agents. In a setting that delivered (e.g., capacity issues, quality issues) as well more closely approximates reality, information is in- as on cash (inventory) management. complete and asymmetrically shared, making price In addition to market mechanisms, pricing in finan- formation a more complex process. The theory of in- cial services may be driven by other factors or centives addresses the informational issues that arise in constraints that may complicate or simplify it. For ex- the principal–agent economic relationship. In such a ample, usury laws specify maximum interest rates to be relationship, a ‘‘principal’’ delegates a set of tasks to an charged for lending to consumers or businesses by ‘‘agent’’ who possesses special competencies to perform banks, credit cards, or pawn shops. Insurance regula- the tasks and the two may have conflicting interests. tions vary by jurisdiction. Fees in open-ended mutual Informational issues present in a principal–agent rela- funds offered to US investors are governed by the In- tionship include: vestment Company Act of 1940. The Employee (i) moral hazard, whereby the agent has private in- Retirement Income Security Act of 1974 (ERISA), which formation that can be used to take actions to regulates US pension plans, specifies that plan trustees serve its interests; such actions may work and investment advisors are considered ‘‘fiduciaries’’ against the interests of the principal, who has who should act in the best interests of plan participants. to assume some of these actions’ adverse con- Among the duties of fiduciaries is to ensure that the sequences, and plan pays reasonable investment expenses, including (ii) adverse selection, where an agent uses private fund, advisory, and custody fees, and trading commis- information about its own characteristics to gain sions. Preferred pricing provisions, known as Most advantage in selecting a contract offered by the Favored Nation (MFN) clauses, are typically included principal. in investment management agreements of pension plans governed by ERISA in the United States and by Non-verifiability is a third issue that may arise when similar laws elsewhere. By the fiduciary standard of the principal and the agent share information that ERISA, transactions in accounts that hold plan assets cannot be verified by a third party, for example, a must reflect the best value for the services received. court of law. The related free-rider problem refers to Such services include execution of trades, research, in- asymmetric sharing of benefits and costs in resource vestment advice, and anything else that may be paid usage among economic agents. The principal–agent through these transaction costs. model addresses the design of contracts with appro- A body of literature exists in economics and finance priate incentives to best align the interests of principal on some aspects of pricing in financial services, and and agent in the presence of the informational issues vendors offer pricing services and software, but little of moral hazard, adverse selection, and non-verifiability. academic research with an operations management Laffont and Martimort (2002) focus on the situation of Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 654 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society a principal dealing with a single agent, to whom the (i) operational infrastructure, which tends to be more principal offers a take-it-or-leave-it contract without sophisticated and expensive for hedge funds negotiation. Their book begins with a review of the than for mutual funds, and has additional com- literature on incentives in economic thought since plexities for SMAs, and Smith (1776) examined incentives in agriculture, and (ii) implementation, whereby profitable opportuni- Hume (1740) explicitly defined the free-rider problem. ties become scarcer as investment vehicles Contract negotiations can be addressed by game the- become larger; and the effectiveness of execution ory; Camerer (2003) wrote a very accessible text on of a vehicle’s investment strategies depends to a behavioral game theory in which he cites all relevant large degree on the market liquidity of the secu- literature. The text by Tirole (1988) contains thorough rities traded by the investment manager in the discussions on price formation. vehicle’s portfolio. Moral hazard and adverse selection are responsible Mutual funds and similarly managed SMAs typi- for anomalies in insurance and credit product pricing. cally charge asset-based fees on a calendar basis. As Akerlof (1970) noted that aggregate risk in segments assets grow due to good performance and inflows of of the insured population will be an increasing func- new funds resulting from this good performance, the tion of insurance premium paid. This is the case manager gets rewarded for skill and effort. As secu- because private knowledge of individuals’ state of rities traded by mutual funds are typically liquid, health, drinking or smoking habits, driving behavior, implementation capacity is rarely an issue. In contrast, stability of employment, etc., leads them to accept hedge funds and like-managed SMAs often face capac- or reject higher premia offered for health, life, auto- ity constraints that limit their size due to diminishing mobile, or mortgage insurance. The same holds true returns to scale, which makes the asset-based fee an for credit, as Stiglitz and Weiss (1981) noted. Higher inadequate incentive for hedge fund managers. Hedge interest rates offered would be a lot more likely to be fund pricing, therefore, constitutes a more representa- accepted by borrowers whose private self-assessment tive application of the principal–agent model because it of their default probability is high. These borrowers must use more complete incentive mechanisms to min- would then become even more likely to default be- imize informational issues between fund manager cause of the high credit cost burden. Insurance and (agent) and investor (principal). A typical pricing struc- credit are therefore rationed, because there is no price ture for hedge funds may consist of high enough to be profitable with certain customers. That means, as Rothschild and Stiglitz (1976) found, (i) a base (or management) fee, which can be a per- that equilibria in competitive insurance markets un- centage of AUM paid on a calendar schedule, der imperfect and asymmetric information needed to and covers operating costs of the fund, and be specified by both price and quantity of contracts (ii) an incentive (or performance) fee, which allows offered. Stiglitz (1977) examined differences in the role the manager to keep a share of the value created of imperfect information in insurance pricing under a for the investor during agreed-on time intervals, monopoly regime compared with perfect competition. to ensure that the interests of the two parties are aligned. 9.3. Pricing in Asset Management, Securities The manager earns an incentive fee if value is cre- Trading and Brokerage, and Credit Cards ated for the investor according to an agreed-on metric Pricing in the asset management industry consists usually based on either monetary units or rates of primarily of fees charged on a client’s assets under return (see Bailey 1990). Typically, incentive fee management (AUM) in investment vehicles such as arrangements have asymmetric payoffs, i.e., they mutual funds, hedge funds, or separately managed reward gains and do not penalize losses. However, accounts (SMAs). Fees can be they often require that an investment’s value be at or (i) fixed, regardless of AUM, above a historical maximum, called a high water mark (ii) asset-based, i.e., a percentage of AUM, or (HWM), before an incentive fee becomes payable to (iii) performance-based, i.e., dependent on AUM’s the manager. This implies that prior losses, if any, return. must have been recovered. Earning high incentive fees depends on fund manager skill; hedge funds with Pricing structures reflect costs of different vehicles, highly skilled managers have higher revenues. Such are formed in the process of bringing investor de- hedge funds can afford the higher expense of setting mand into equilibrium with each vehicle’s capacity, up and maintaining a robust operational infrastruc- and attempt to address the principal–agent issues be- ture and control framework, as discussed by Alptuna tween investor (principal) and investment manager et al. (2010). Incentive fees are also common in private (agent). Capacity of an investment vehicle refers to: equity, where they are typically known as carried Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 655 interest and paid by investors on the liquidation of a structure requires a very high capital investment and partnership and distribution of proceeds. is costly to operate. Generally, institutions have access A new rule adopted by the Securities and Exchange to lower pricing than individual investors. A pricing Commission under Section 205 of the Investment application for brokerage commissions and invest- Adviser’s Act of 1940 permitted the use of perfor- ment advisory fees was studied by Altschuler et al. mance fees by registered investment advisers. Several (2002). They developed models to determine appro- published articles then examined performance fee priate commission rates for a new discount brokerage schemes, with emphasis on inherent moral hazard channel, and asset-based fees for advice and unlim- and its mitigation. Record and Tynan (1987) described ited trading in full-service accounts introduced by incentive fee arrangements, and examined the basic Merrill Lynch. Among the issues to contend with were issues involved. Davanzo and Nesbitt (1987) analyzed adverse selection, which might prompt full-service incentive fee structures and their impact to the busi- clients without a strong relationship with their finan- ness of investment management. Kritzman (1987) cial advisors to select the low-cost discount channel. proposed ways to deal with moral hazard issues, Another known issue is moral hazard, when an ad- and to reward the manager’s skill rather than invest- visor might not be committing much effort to a client’s ment style or chance; these issues were also addressed portfolio after the account was converted to the asset- by Grinblatt and Titman (1987). Grinold and Rudd based annual fee. (1987) added another perspective by examining fee Credit card pricing and line management were structures that would be appropriate for the investor studied together by Trench et al. (2003). They built a and for the manager according to the latter’s invest- Markovian model to select the optimal APR and credit ment skills. Bailey (1990) demonstrated that incentive line for each individual cardholder based on historical fee metrics based on value added in monetary units behavior with the goal to maximize Bank One’s serve investors’ interests better than metrics based on profitability. Offering an attractive APR and a large rates of return. Lynch and Musto (1997) examined credit line to entice a cardholder to transfer a balance how incentive fees compare with asset-based fees, presented the moral hazard issue of the client taking especially in extracting value-creating effort by the advantage of the offer and then defaulting, which the manager. Anson (2001) valued the call option implicit model successfully addressed. In his book, Phillips in incentive fees by Black–Scholes analysis. The non- (2005) discusses consumer loan and insurance pricing. linear optionality in incentive fees was found by He expands on it in a more recent book (Phillips 2010) Li and Tiwari (2009) to be optimal even for mutual that also includes a chapter by Caufield (2010), who funds. Golec and Starks (2004) examined the reduc- examines pricing for consumer credit, including credit tion in risk levels of mutual funds after the option-like cards, mortgages, and auto loans. In his book, Thomas incentive fee option was prohibited by an act of Con- (2009) discusses the use of risk-based pricing in con- gress in 1971. sumer credit. Wuebker et al. (2008) also wrote a book More recent works have studied the role of HWMs on pricing in financial services. in incentive fee contracts. In a seminal paper, Goetz- mann et al. (2003) valued incentive fee contracts with HWMs using models with closed-form solutions. An- 9.4. Revenue Management in Financial Services: son (2001) and Lee et al. (2004) examine the free-rider Challenges and Opportunities problem in hedge funds that offer all investors the Revenue management principles can be applied to same HWM: late investors can avoid paying incentive financial services pricing with some adaptation. The fees if they enter after losses suffered by earlier in- framework can address key considerations, such as vestors. This issue was addressed in practice by rationing in credit and insurance (see, e.g., Akerlof offering investors a different HWM based on their 1970, Phillips 2010, Stiglitz and Weiss 1981). Evidence time of entry. of rationing is reflected in protecting capacity for air- Pricing for security transactions is determined in a line fare classes (Talluri and Van Ryzin 2005). Another deregulated and competitive market. It has been ex- key consideration is consumer behavior (Talluri periencing a downward trend driven by the dramatic and Van Ryzin 2004). Some revenue management cost reductions brought by technological innovations concepts, for example, capacity, can be quite different such as electronic trading (see Bortoli et al. 2004, in financial services than in the transportation and Levecq and Weber 2002, Stoll 2006, 2008, Weber 1999, hospitality industries. It can also be idiosyncratic to 2006). Transaction commissions typically pay for the each financial product, for example, a CD vs. a mutual costs of brokerage, clearing and execution, trading, fund, or a hedge fund. Boyd (2008) discussed the research, and investment advice before the brokerage challenges faced by pricing researchers. Identification firm can make a profit. Economies of scale are very problems exist (see Koopmans 1949, Manski 1995) that important, because a brokerage’s operational infra- make it hard to build demand curves from data and Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 656 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society therefore figure out demand elasticities, which are key evidence suggests that it is six times more expensive to pricing. These problems are especially acute in in- to acquire a new client than to service an existing one, dustries where sales and client relationship persons making operations a really important factor in finan- have a lot of information on customers, including de- cial services. mand and price elasticities. Such information remains As described in this paper, there is an extensive lit- private and not centrally shared, as is often the case in erature on traditional service operations research some areas of financial services. For example, firms topics such as waiting lines, forecasting, and person- have databases with prices of closed sales but no in- nel scheduling that are applicable to financial services formation on prices of refused offers; the latter as well. Inventory models have been successfully ap- remains in the field. Remedies have been proposed plied to cash and currency management. Operational and can be considered, such as Gonik’s (1978), which risk management is an emerging area that is attracting Chen (2005) analyzed and compared with a set of lin- quite a lot of attention lately. We expect researchers to ear contracts. Tools used by Athey and Haile (2002) branch out and address other non-traditional opera- may also be considered. tional issues in financial services, some of which we have highlighted here. Many of these are likely to be 10. Concluding Remarks and Future cross-disciplinary, interfacing information technology, marketing, finance, and statistics. Research Directions Another area with a potentially significant payoff is In this paper, we have attempted to present an over- the optimization of execution costs in securities trad- view of operations management in the financial ing. Amihud and Mendelson (2010) and Goldstein services industry, and tried to make the case that this et al. (2009) examine transaction costs in recent studies, industry has several unique characteristics that de- and Bertsimas and Lo (1998) develop trading strate- mand attention separate from research in services in gies that optimize the execution of equity transactions. general. We have identified a number of specific char- A significant body of work exists in algorithmic trading, acteristics that make financial services unique as far as market microstructure, and the search for liquidity in product design and service delivery are concerned, securities markets (see, e.g., O’Hara 1998). Operations requiring an interdisciplinary approach. In Appendix management researchers can examine the problem and A, we provide an overview table of the various op- develop solutions by synthesizing elements from this erational processes in financial services and highlight diverse set of disciplines and points of view. the ones that have attracted attention in operations Another interesting area of research could be the management literature. From the table in Appendix integration of the various objectives for improving A, it becomes immediately clear that many processes operations in financial services in which interactions in the financial services industries have received scant among components can be viewed and modeled research attention from the operational point of view holistically. For example, most financial services are and that there are several areas that are worthy of quite keen on improving the productivity of their pro- research efforts in the future. These include each step cesses. However, one has to keep in mind that there are in the financial product and service life cycle as well strong relationships between the productivity of the as in the customer relationship life cycle. processes, the operational risk encountered in these Much work remains to be done on the design of processes, and the quality of the services delivered to financial products so that they are the clients. When one reduces headcount, productivity may indeed go up; however, the operational risk may (i) easier to understand for the customer (result- increase and the quality of service may go down. It is of ing in fewer calls to call centers), great interest to the financial services industry that (ii) easier to use (better online and face-to-face these interdependencies are well understood. interactions, with less waiting), Pricing and revenue management of financial (iii) less prone to operational risks induced by services could be an area that is ripe for academic human errors, research with a potential short-term payoff that may (iv) easier to forecast and arrange the necessary be large. Operations management researchers could operational resources for, and leverage related work in economics, finance, and may (v) able to take advantage of pricing and revenue adapt revenue management principles to develop management opportunities. novel pricing methodologies for financial services. Service designs need to recognize the fact that Operations management research may also be key financial services are relatively sticky, involve in enabling visionary ideas for reforming corporate long-term relationships with customers, and are at governance. In the principal–agent relationship be- the same time prone to attrition due to poor perfor- tween corporate shareholders and management, mance or frustrating service encounters. Anecdotal proxy voting is the most important tool available to Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 657 shareholders for ensuring that management acts of proxy aggregation mechanisms is almost certainly according to their interests. This tool is seldom effec- expected to face complex logistical issues, which op- tive due to shareholder apathy and obstacles to voting erations management can examine and address. through institutional ownership of shares. Holton (2006) proposed proxy aggregation mechanisms, such Acknowledgments as a proxy exchange, that could make the process The authors thank five anonymous referees, an editor, and more effective in improving corporate governance. other readers for constructive comments that helped revise Such mechanisms would ensure that the voting of an earlier version of the paper. The presentation, content, proxies is handled by informed entities acting in the flow, and readability of the paper has benefited as a result. best interests of shareholders, with thorough knowl- The views expressed in this paper are those of the authors edge of the issues to be voted on. Implementation and do not necessarily reflect those of Goldman Sachs.

Appendix A: Operations Management Processes in Financial Services

Operational processes Strategic processes

Current customer portfolio Delinquent customer Service/process Process realm Acquisition/origination management management Product design design Retail banking Campaign management Cash, currency, and deposit Collections Account type Branch location and management offerings and design rationalization Deposit product operations Teller, ATM, and online operations Recovery operations Outsourcing / insourcing Retail loan origination Treasury operations ATM location planning Risk management Capacity issues Credit, liquidity, operational Technology issues Anti-money laundering surveillance Commercial Loan syndication Credit agreement maintenance Defaulted borrower Credit facility Credit facility design lending management offerings Lending operations Loan syndicate management Recovery operations Revolver, term, Syndicated loan offerings bridge, letter of credit, etc. Credit risk management Risk management Outsourcing / insourcing Credit, liquidity, Capacity issues operational Technology issues Insurance Acquisition planning Policy maintenance Past due account Product design Product offerings handling Solicitation management Claims investigation and Recovery operations Insurance policy Outsourcing / insourcing processing types Application processing Worker’s comp case tracking Annuity products Capacity issues Underwriting Property disposal Technology issues (auto auctions, etc.) Risk management Operational Credit cards Campaign creation and Statement operations Collections Card offerings Facilities location management management Credit risk management Remittance processing Recovery operations APR, credit line, Processing and call rewards centers Plastic printing and Call center management Merchant acquisition Outsourcing/insourcing mailing operations operations Plastic printing and Private label cards Capacity issues mailing operations Risk management Technology issues Credit, operational

Continued Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 658 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

Appendix A (Continued)

Operational processes Strategic processes

Current customer portfolio Delinquent customer Service/process Process realm Acquisition/origination management management Product design design Mortgage Mortgage origination Mortgage servicing operations Collections Product offerings Facilities location banking Securitization Remittance processing Foreclosure management Term, APR Processing and call centers Secondary market Call center management Recovery operations Outsourcing/insourcing operations Credit risk management Risk management Capacity issues Credit, operational Technology issues Brokerage/ Client prospecting Client at risk handling Collections operations Account offerings Multichannel design and investment management pricing advisory Cold calling, referrals Client litigation Delinquent margin accounts Outsourcing/insourcing Anti-money laundering Commission management Recovery operations Capacity issues Margin lending Technology issues Trust services Trading platform design Channel choice and coordination Risk management Credit, operational Anti-money laundering Asset Investment manager Pricing structures Collections operations New product/fund Organization design management due diligence design Pricing structures Custodian operations Recovery operations Share class/series Location of sales and management client relationship management New account setup Fund administration Outsourcing/insourcing processes Operational risk Fund accounting Capacity issues management Anti-money laundering Valuation operations Infrastructure, implementation Operational risk management Technology issues Anti-money laundering

Processes in bold have been addressed in operations management literature. Processes in italics have received less attention in the operations management literature.

References agement—Processes and Costs. Palgrave Macmillan, New York, Akerlof, G. A. 1970. The market for ‘‘lemons’’: Quality uncertainty 18–41. and the market mechanism. Q. J. Econ. 84(3): 488–500. Alternative Investment Management Association. 2007. Guide to Akerlof, G. A., R. D. Milbourne. 1978. New calculations of income sound practices for European hedge fund managers. Available at and interest elasticities in Tobin’s model of the transactions de- http://www.aima.org/en/knowledge_centre/sound-practices/ mand for money. Rev. Econ. Stat. 60(4): 541–546. guides-to-sound-practices.cfm (accessed date August 3, 2009). Aksin, O. Z., M. Armony, V. Mehrotra. 2007. The modern call-center: Altschuler, S., D. Batavia, J. Bennett, R. Labe, B. Liao, R. Nigam, J. Oh. A multi-disciplinary perspective on operations management 2002. Pricing analysis for Merrill Lynch Integrated Choice. Inter- research. Prod. Oper. Manag. 16(6): 665–688. faces 32(1): 5–19. Aldor-Noiman, S., P. D. Feigin, A. Mandelbaum. 2009. Workload Amihud, Y., H. Mendelson. 2010. Transaction costs and asset forecasting for a call center methodology and a case study. Ann. management. Pinedo, M. ed. Operational Control in Asset Appl. Stat. 3(4): 1403–1447. Management—Processes and Costs. Palgrave Macmillan, Alptuna, U¨ ., M. Hatzakis, R. Tu¨tu¨ncu¨ . 2010. A best practices New York, 170–189. framework for operational infrastructure and controls in asset Andrews, B. H., S. M. Cunningham. 1995. L. L. Bean improves call- management. Pinedo, M. ed. Operational Control in Asset Man- center forecasting. Interfaces 25(6): 1–13. Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 659

Anson, M. J. P. 2001. Hedge fund incentive fees and the ‘free option.’ Bensoussan, A., A. Chutani, S. P. Sethi. 2009. Optimal cash man- J. Altern. Invest. 4(2): 43–48. agement under uncertainty. Oper. Res. Lett. 37(6): 425–429. Antelman, G. R., I. R. Savage. 1965. Surveillance problems: Wiener Berger, A. N., D. D. Humphrey. 1997. Efficiency of financial insti- process. Nav. Res. Logist. Q. 12(1): 35–55. tutions: International survey and directions for future research. Antipov, A., N. Meade. 2002. Forecasting call frequency at a finan- Eur. J. Oper. Res. 98(2): 175–212. cial services call centre. J. Oper. Res. Soc. 53(9): 953–960. Bertsimas, D., A. Lo. 1998. Optimal control of execution costs. Apte, A., U. M. Apte, R. P. Beatty, I. C. Sarkar, J. H. Semple. 2004. J. Financ. Mark. 1: 1–50. The impact of check sequencing on NSF (not-sufficient funds) Bhulai, S., G. Koole, A. Pot. 2008. Simple methods for shift sched- fees. Interfaces 34(2): 97–105. uling in multiskill call centers. Manuf. Serv. Oper. Manage. 10(3): Apte, U., R. A. Cavaliere, S. S. Kulkarni. 2010. Analysis and im- 411–420. provement of information intensive services: Evidence from Biggs, J. 2010. Management of risk, technology and costs in a multi- insurance claims handling. Prod. Oper. Manag. 19(6): 665–678. line asset management business. Pinedo, M. ed. Operational Apte, U., C. Maglaras, M. Pinedo. 2008. Operations in the service Control in Asset Management—Processes and Costs. Palgrave industries: Introduction to the special issue. Prod. Oper. Manag. Macmillan, New York, 88–107. 17(3): 235–237. Bitran, G. R., J.-C. Ferrer, P. Rocha e Oliviera. 2008. Managing cus- Arfelt, K. 2010. Lean six sigma in asset management: A way to cut tomer experiences: Perspectives on the temporal aspects of costs? Pinedo, M. ed. Operational Control in Asset Management— service encounters. Manuf. Serv. Oper. Manage. 10(1): 61–83. Processes and Costs. Palgrave Macmillan, New York, 60–87. Black, K. H. 2007. Preventing and detecting hedge fund failure risk Armony, M., C. Maglaras. 2004. Contact centers with a call-back op- through partial transparency. Derivatives Use, Trading and Reg- tion and real-time delay information. Oper. Res. 52(4): 527–545. ulation 12(4): 330–341. Arvedlund, E. 2009. Too Good to be True: The Rise and Fall of Bernie Boeschoten, W. C., M. M. G. Fase. 1992. The demand for large bank Madoff. Portfolio Hardcover. Penguin Publishing, New York. notes. J. Money, Credit Bank. 24(3): 319–337. Asset Managers’ Committee to the President’s Working Group on Fi- Bohn, J., D. Hancock, P. W. Bauer. 2001. Estimates of scale and cost nancial Markets. 2009. Best practices for the hedge fund industry. efficiency for Federal Reserve currency operations. Econ. Rev. Available at http://www.amaicmte.org/Public/AMC%20Report% (Federal Reserve Bank of Cleveland) 37(4): 2–26. 20-%20Final.pdf (accessed date August 3, 2009). Bollapragada, S., S. K. Nair. 2010. Improving right party contact Athanassopoulos, A. D., D. Giokas. 2000. The use of data envelopment rates at outbound call centers. Prod. Oper. Manag. 19(6): 769–779. analysis in banking institutions: Evidence from the Commercial Bortoli, L. G., A. Frino, E. Jarnecic. 2004. Differences in the cost of Bank of Greece. Interfaces 30(2): 81–95. trade execution services on floor-based and electronic futures Athey, S., P. A. Haile. 2002. Identification of standard auction markets. J. Financ. Serv. Res. 26(1): 73–87. models. Econometrica 70(6): 2107–2140. Boyd, E. A. 2008. Challenges faced by researchers in pricing. Lecture Avramidis, A. N., A. Deslauriers, P. L’Ecuyer. 2004. Modeling daily series of the Center for Analytical Research in Technology arrivals to a telephone call center. Manage. Sci. 50(7): 896–908. (CART), Tepper School of Business, Carnegie Mellon Univer- sity, February 5. Available at http://mat.tepper.cmu.edu/blog/ Bagchi, K., G. Udo. 2003. An analysis of the growth of computer and ?p=230 (accessed date April 25, 2010). internet security breaches. Comm. of AIS 12: 684–700. Boyer, K. K., R. Hallowell, A. V. Roth. 2002. E-services operating Bailey, J. V. 1990. Some thoughts on performance-based fees. Financ. strategy—a case study and a method for analyzing operational Anal. J. 46(4): 31–40. benefits. J. Oper. Manage. 20(2): 175–188. Balasubramanian, S., P. Konana, N. M. Menon. 2003. Customer sat- Brazier, F. M. T., C. M. Jonker, F. J. Jungen, J. Treur. 1999. Distributed isfaction in virtual environments: A study of online investing. scheduling to support a call centre: A co-operative multi-agent Manage. Sci. 49(7): 871–889. approach. Appl. Artif. Intell. J. 13, Special Issue on Multi-Agent Barlow, R., F. Proschan. 1975. Statistical Theory of Reliability and Life Systems. H. S. Nwana and D. T. Ndumu (eds.), 65–90. Testing—Probability Models. Holt, Rinehart and Winston, New Brown, L., N. Gans, A. Mandelbaum, A. Sakov, H. Shen, S. Zeltyn, L. York. Zhao. 2005. Statistical analysis of a telephone call center: Barth, W., M. Manitz, R. Stolletz. 2010. Analysis of two-level support A queueing-science perspective. J. Am. Stat. Assoc. 100(469): systems with time-dependent overflow—a banking application. 36–50. Prod. Oper. Manag. 19(6): 757–768. Brown, S. J., W. N. Goetzmann, B. Liang, C. Schwartz. 2009a. Es- Basel Committee. 2003. Sound practices for the management and timating operational risk for hedge funds: The o score. Financ. supervision of operational risk. Bank for International Settle- Anal. J. 65(1): 43–53. ments, Basel Committee Publications No. 96. Brown, S. J., W. N. Goetzmann, B. Liang, C. Schwartz. 2009b. Trust Baesens, B., T. Van Gestel, S. Viaene, M. Stepanova, J. Suykens, J. and delegation. NBER Working Paper No. w15529. Available at Vanthienen. 2003. Benchmarking state-of-the-art classification http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1510479 algorithms for credit scoring. J. Oper. Res. Soc. 54(6): 627–635. (accessed date September 4, 2010). Bassamboo, A., J. M. Harrison, A. Zeevi. 2005. Dynamic routing and Buell, R., D. Campbell, F. Frei. 2010. Are self-service customers sat- admission control in high-volume service systems: Asymptotic isfied or stuck? Prod. Oper. Manag. 19(6): 679–697. analysis via multi-scale fluid limits. J. Queueing Syst. 51(3–4): Camerer, C. F. 2003. Behavioral Game Theory: Experiments in Strategic 249–285. Interaction. Princeton University Press, Princeton, NJ. Bauer, P. W., A. Burnetas, V. CVSA, G. Reynolds. 2000. Optimal Campbell, D., F. Frei. 2010a. Cost structure patterns in the asset use of scale economies in the Federal Reserve’s currency infra- management industry. Pinedo, M. ed. Operational Control in structure. Econ. Rev. (Federal Reserve Bank of Cleveland) 36(3): Asset Management—Processes and Costs. Palgrave Macmillan, 13–27. New York, 154–168. Bather, I. A. 1966. A continuous time inventory model. J. App. Prob. Campbell, D., F. Frei. 2010b. Cost structure, customer profitability, 3: 538–549. and retention implications of self-service distribution channels: Baumol, W. J. 1952. The transactions demand for cash: An inventory Evidence from customer behavior in an online banking channel. theoretic approach. Q. J. Econ. 66(4): 545–556. Manage. Sci. 56(1): 4–24. Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 660 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

Capon, N. 1982. Credit scoring systems: A critical analysis. J. Mark. Daellenbach, H. G. 1971. A stochastic cash balance model with two 46(2): 82–91. sources of short-term funds. Inter. Econ. Rev. 12(2): 250–256. Caufield, S. 2010. Pricing in consumer credit. O¨ zer, O¨ ., R. Phillips, Davanzo, L. E., S. L. Nesbitt. 1987. Performance fees for investment eds. Handbook of Pricing Management. Oxford University management. Financ. Anal. J. 43(1): 14–20. Press, Oxford, UK, to appear. Dawande, M., M. Mehrotra, V. Mookerjee, C. Sriskandarajah. 2010. CFA Institute. 2009. Asset Manager Code of Professional Conduct. 2nd An analysis of coordination mechanisms for the U.S. cash sup- edn Available at http://www.cfapubs.org/doi/pdf/10.2469/ ply chain. Manage. Sci. 56(3): 553–570. ccb.v2009.n8.1 (acceded date August 3, 2009). De Almeida Filho, A. T., C. Mues, L. C. Thomas. 2010. Optimizing Chan, P. K., W. Fan, A. L. Prodromidis, S. J. Stolfo. 1999. Distributed the collections process in consumer credit. Prod. Oper. Manag. data mining in credit card fraud detection. IEEE Intell. Syst. 19(6): 698–708. 14(6): 67–74. De Haan, L., A. Ferreira. 2006. Extreme Value Theory—An Introduc- Chen, B. P. K., S. G. Henderson. 2001. Two issues in setting call tion. Springer, New York. centre staffing levels. Ann. Oper. Res. 108(1–4): 175–192. de Lascurain, M., L. de los Santos, F. J. Herrerı´a, D. F. Mun˜oz, A. Chen, F. 2005. Salesforce incentives, market information, and pro- Palacios-Brun, O. Romero-Herna´ndez, F. Solı´s. 2011. INDEVAL duction/inventory planning. Manage. Sci. 51(1): 60–75. develops a new operating and settlement system using oper- Chen, P. Y., L. M. Hitt. 2002. Measuring switching costs and the ations research. Working paper. Instituto para el Deposito de determinants of customer retention in internet enabled busi- Valores, Mexico City, Mexico, to appear in Interfaces. January/ nesses: A study of the online brokerage industry. Infor. Syst. Res. February 2011. 13(3): 255–274. Dekker, R., M. Fleischmann, K. Inderfurth, L. N. Van Wassenhove, Cheng, F., N. Jengte, W. Min, B. Ramachandran, D. Gamarnik. 2007. eds. 2004. Reverse Logistics: Quantitative Methods for Closed-Loop Modeling operational risk in business processes. J. Oper. Risk Supply Chains. Springer-Verlag, Berlin, Germany. 2(2): 73–98. Dhillon, G. 2004. The challenge of managing information security. Chernobai, A. S., S. T. Rachev, F. J. Fabozzi. 2007. Operational Int. J. Inf. Manage. 24(1): 3–4. Risk—A Guide to Basel II Capital Requirements. Models and Dijk, N. M., van, E. van der Sluis. 2008. To pool or not to pool in call Analysis. The Frank J. Fabozzi Series. John Wiley and Sons, centers. Prod. Oper. Manag. 17(3): 296–305. New York. Duffy, T., M. Hatzakis, W. Hsu, R. Labe, B. Liao, X. Luo, J. Oh, A. Chevalier, P., J.-C. Van den Schrieck. 2008. Optimizing the staffing Setya, L. Yang. 2005. Merrill Lynch improves liquidity risk and routing of small-size hierarchical call centers. Prod. Oper. management for revolving credit lines. Interfaces 35(5): 353–369. Manag. 17(3): 306–319. Eisenberg, L., T. H. Noe. 2001. Systemic risk in financial systems. Ciciretti, R., I. Hasan, C. Zazzara. 2009. Do internet activities add Manage. Sci. 47(2): 236–249. value? Evidence from the traditional banks. J. Financ. Serv. Res. Elsinger, H., A. Lehar, M. Summer. 2006. Risk assessment for bank- 35(1): 81–98. ing systems. Manage. Sci. 52(9): 1301–1314. Clemons, E. K., L. M. Hitt, B. Gu, M. E. Thatcher, B. W. Weber. 2002. Eppen, G. D., E. F. Fama. 1968. Solutions for cash-balance and sim- Impacts of e-commerce and enhanced information endowments ple dynamic-portfolio problems. J. Bus. 41(1): 94–112. on financial services: A quantitative analysis of transparency, differential pricing, and disintermediation. J. Financ. Serv. Res. Eppen, G. D., E. F. Fama. 1969. Cash balance and simple dynamic 22(1–2): 73–90. portfolio problems with proportional costs. Int. Econ. Rev. 10(2): 119–133. Constantinides, G. M. 1976. Stochastic cash management with fixed and proportional transaction costs. Manage. Sci. 22(12): Eppen, G. D., E. F. Fama. 1971. Three asset cash balance and dy- 1320–1331. namic portfolio problems. Manage. Sci. 17(5): 311–319. Constantinides, G. M., S. F. Richard. 1978. Existence of simple pol- Fase, M. M. G. 1981. Forecasting the demand for banknotes: Some icies for discounted-cost inventory and cash management in empirical results for the Netherlands. Eur. J. Oper. Res. 6(3): 269– continuous time. Oper. Res. 26(4): 620–636. 278. Cook, W. D., L. M. Seiford. 2009. Data envelopment analysis Fase, M. M. G., D. Van der Hoeven, M. Van Nieuwkerk. 1979. A (DEA)—thirty years on. Eur. J. Oper. Res. 192(1): 1–17. numerical planning model for a central bank’s bank notes op- erations. Stat. Neerl. 33(1): 7–25. Cooper, W. W., L. M. Seiford, K. Tone. 2007. Data Envelopment Analysis: A Comprehensive Text with Models, Applications, Feldman, Z., A. Mandelbaum, W. A. Massey, W. Whitt. 2008. References and DEA-Solver Software. 2nd edn. Springer, New Staffing of time varying queues to achieve time-stable perfor- York. mance. Manage. Sci. 54(2): 324–338. Cruz, M. 2002. Modeling, Measuring and Hedging Operational Risk. Fiterman, A. E., V. G. Timkovsky. 2001. Basket problems in margin John Wiley and Sons, New York. calculation: Modelling and algorithms. Eur. J. Oper. Res. 129(1): 209–223. Cruz, M. 2010. Strategic and tactical cost management in the asset management industry. Pinedo, M. ed. Operational Control in Frei, F. X. 2006. Breaking the trade-off between efficiency and Asset Management—Processes and Costs. Palgrave Macmillan, service. Harv. Bus. Rev. 84(11): 92–101. New York, 108–131. Frei, F. X., R. Kalakota, A. J. Leone, L. M. Marx. 1999. Process vari- Cruz, M., M. Pinedo. 2008. Total quality management and opera- ation as a determinant of bank performance: Evidence from the tional risk in the service industries. Chen, Z.-L., S. Raghavan, retail banking industry. Manage. Sci. 45(9): 1210–1220. eds. Chapter 7 in Tutorials in Operations Research 2008. INFORMS, Fung, M. K. 2008. To what extent are labor-saving technologies im- Hanover, MD, 154–169. proving efficiency in the use of human resources? Evidence from Cruz, M., M. L. Pinedo. 2009. Global asset management: Costs, the banking industry. Prod. Oper. Manag. 17(1): 75–92. productivity, and operational risk. J. Appl. IT Invest. Manage. Furst, K., W. W. Lang, D. E. Nolle. 2002. Internet banking. J. Financ. 1(2): 5–12. Serv. Res. 22(1–2): 95–117. Cummins, J. D., M. A. Weiss, H. Zi. 1999. Organizational form and Gans, N., G. Koole, A. Mandelbaum. 2003. Telephone call centers efficiency: The coexistence of stock and mutual property-liabil- tutorial. Review and research prospects. Manuf. Serv. Oper. ity insurers. Manage. Sci. 45(9): 1254–1269. Manage. 5(2): 79–141. Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 661

Gans, N., H. Shen, Y.-P. Zhou, N. Korolev, A. McCord, H. Ristock. Heskett, J. L., T. O. Jones, G. W. Loveman, W. E. Sasser, L. A. Schle- 2009. Parametric stochastic programming models for call-center singer. 1994. Putting the service-profit chain to work. Harv. Bus. workforce scheduling. Available at http://faculty.washington. Rev. 72(2): 164–174. edu/yongpin/Stochastic-Workforce-Scheduling.pdf (accessed Hitt, L. M., F. X. Frei. 2002. Do better customers utilize electronic date August 22, 2010). distribution channels? The case of PC banking. Manage. Sci. Gans, N., Y.-P. Zhou. 2003. A call-routing problem with service-level 48(6): 732–748. constraints. Oper. Res. 51(2): 255–271. Holton, G. A. 2006. Investor suffrage movement. Financ. Anal. J. Garg, A., J. Curtis, H. Halper. 2003. Quantifying the financial impact 62(6): 15–20. of IT security breaches. Inf. Manage. Comput. Secur. 11(2): 74–83. Hong, K.-S., Y.-P. Chi, L. R. Chao, J.-H. Tang. 2003. An integrated Gawande, M. 1996. Workforce scheduling problems with side con- system theory of information security management. Inf. Man- straints. Presentation at the semi-annual INFORMS meeting in age. Comput. Secur. 11(5): 243–248. Washington, DC. Hume, D. A. 1740. A Treatise of Human Nature. Oxford University Geismar, N., M. Dawande, D. Rajamani, C. Sriskandarajah. 2007. Press (2000 Edition). Managing a bank’s currency inventory under new Federal Re- Humphrey, D. B., L. B. Pulley, J. M. Vesala. 2000. The check’s in the serve guidelines. Manuf. Serv. Oper. Manage. 9(2): 147–167. mail: Why the United States lags in the adoption of cost-saving Giloni, A., S. Seshadri, P. V. Kamesam. 2003. Service system design electronic payments. J. Financ. Serv. Res. 17(1): 17–19. for the property and casualty insurance industry. Prod. Oper. Janosi, T., R. Jarrow, F. Zullo. 1999. An empirical analysis of the Manag. 12(1): 62–78. Jarrow-van Deventer model for valuing non-maturity demand Girgis, N. M. 1968. Optimal cash balance levels. Manage. Sci. 15(3): deposits. J. Deriv. 7(1): 8–31. 130–140. Jarrow, R. A. 2008. Operational risk. J. Bank. Finance 32(5): 870–879. Goetzmann, W. N., J. E. Jr. Ingersoll, S. A. Ross. 2003. High-water Jarrow, R. A., J. Oxman, Y. Yildirim. 2010. The cost of operational marks and hedge fund management contracts. J. Finance 58(4): risk loss insurance. Review of Derivatives Research. 13(3): 273–295. 1685–1718. Jarrow, R. A., D. R. van Deventer. 1998. The arbitrage-free valuation Goldstein, M. A., P. Irvine, E. Kandel, Z. Wiener. 2009. Brokerage and hedging of demand deposits and credit card loans. J. Bank. commissions and institutional trading patterns. Rev. Financ. Finance 22(3): 249–272. Stud. 22(12): 5175–5212. Jewell, W. S. 1974. Operations research in the insurance industry: 1. Golec, J., L. Starks. 2004. Performance fee contract change and mu- A survey of applications. Oper. Res. 22(5): 918–928. tual fund risk. J. Financ. Econ. 73(1): 93–118. Jones, P. 1988. Quality, capacity and productivity in service indus- Gonik, J. 1978. Tie salesmen’s bonuses to their forecasts. Harv. Bus. tries. Int. J. Hosp. Manage. 7(2): 104–112. Rev. 56: 116–123. Jongbloed, G., G. M. Koole. 2001. Managing uncertainty in call cen- Green, L., P. Kolesar, W. Whitt. 2007. Coping with time-varying de- ters using Poisson mixtures. Appl. Stoch. Models Bus. Ind. 17(4): mand when setting staffing requirements for a service system. 307–318. Prod. Oper. Manag. 16(1): 13–39. Kantor, J., S. Maital. 1999. Measuring efficiency by product group: Grinblatt, M., S. Titman. 1987. How clients can win the gaming Integrating DEA with activity-based accounting in a large Mid- game. J. Portfolio Manage. 13(4): 14–20. east bank. Interfaces 29(3): 27–36. Grinold, R., A. Rudd. 1987. Incentive fees: Who wins? Who loses? Katz, K. L., B. M. Larson, R. C. Larson. 1991. Prescription for the Financ. Anal. J. 43(1): 27–38. waiting-in-line blues: Entertain, enlighten, and engage. Sloan Gu¨ntzer, M. M., D. Jungnickel, M. Leclerc. 1998. Efficient algorithms Manage. Rev. 32: 44–53. for the clearing of interbank payments. Eur. J. Oper. Res. 106(1): Kekre, S., N. Secomandi, E. So¨nmez, K. West. 2009. Balancing risk 212–219. and efficiency at a major commercial bank. Manuf. Serv. Oper. Gurvich, I., J. Luedtke, T. Tezcan. 2010. Staffing call centers with Manage. 11(1): 160–173. uncertain demand forecasts: A chance-constrained optimiza- Kleinrock, L. 1976. Queueing Systems—Volume 2: Computer Applica- 56 tion approach. Manage Sci. (7): 1093–1115. tions. Wiley Interscience, New York. Halfin, S., W. Whitt. 1981. Heavy-traffic limits for queues with many Koetter, M. 2006. Measurement matters - Alternative input price 29 exponential servers. Oper. Res. (3): 567–588. proxies for bank efficiency analyses. J. Financ. Serv. Res. 30(2): Hammond, D., S. Mahesh. 1995. A simulation and analysis of bank 199–227. teller manning. Proceedings of the 1995 Winter Simulation Confer- Kolesar, P. 1984. Stalking the endangered CAT: A Queueing analysis of ence, 1077–1080. congestion at automatic teller machines. Interfaces 14(6): 16–26. Hand, D. J., W. E. Henley. 1997. Statistical classification methods in Koopmans, T. C. 1949. Identification problems in economic model 160 consumer credit scoring: A review. J. R. Stat. Soc. (3): 523– construction. Econometrica 17(2): 125–144. 541. Krishnan, M. S., V. Ramaswamy, M. C. Meyer, P. Damien. 1999. Harker, P. T., S. A. Zenios. 1999. Performance of financial institu- Customer satisfaction for financial services: The role of prod- tions. Manage. Sci. 45(9): 1175–1176. ucts services and information technology. Manage. Sci. 45(9): Harker,P.T.,S.A.Zenios,eds.2000.Performance of Financial Institu- 1194–1209. tions—Efficiency, Innovation, Regulation. Cambridge University Kritzman, M. P. 1987. Incentive fees: Some problems and some Press, Cambridge, UK. solutions. Financ. Anal. J. 43(1): 21–26. Harrison, J. M., A. Zeevi. 2005. A method for staffing large call Kundro, C., S. Feffer. 2003. Understanding and mitigating opera- centers based on stochastic fluid models. Manuf. Serv. Oper. tional risk in hedge fund investments. Capco Institute, White 7 Manage. (1): 20–36. paper series. Available at http://www.capco.com/content/ Hausman, W. H., A. Sanchez-Bell. 1975. The stochastic cash balance knowledge-ideas?q=content/research (accessed date August problem with average compensating-balance requirements. 31, 2009). 21 Manage. Sci. (8): 849–857. Kundro, C., S. Feffer. 2004. Valuation issues and operational risk in Henderson, S. G. 2003. Estimation for nonhomogeneous Poisson hedge funds. Capco Institute, Working paper. Available at http:// processes from aggregated data. Oper. Res. Lett. 31(5): 375–382. www.capco.com/files/pdf/74/02_RISKS/04_Valuation%20issues Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 662 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

%20and%20operational%20risk%20in%20hedge%20funds.pdf Mehrotra, V., K. Ross, G. Ryder, Y.-P. Zhou. 2009. Routing to manage (accessed date August 31, 2009). resolution and waiting time in call centers with heterogeneous Labe, R. 1994. Database marketing increases prospecting effective- servers. Available at http: //faculty.washington.edu/yongpin/ ness at Merrill Lynch. Interfaces 24(5): 1–12. wait_and_resolution.pdf (accessed date August 22, 2010). Labe, R., I. Papadakis. 2010. Propensity score matching applied to Melnick, E. L., P. R. Nayyar, M. L. Pinedo, S. Seshadri. 2000. Creating financial service analytics. Bank of America Working paper. Value in Financial Services—Strategies, Operations and Technolo- gies. Kluwer Academic Publishers, Norwell, MA. Ladany, S. P. 1997. Optimal banknote replenishing policy by an issuing bank. Int. Trans. Oper. Res. 4(1): 1–12. Menor, L. J., A. V. Roth. 2008. New service development competence and performance: An empirical investigation in retail banking. Laffont, J.-J., D. M. Martimort. 2002. The Theory of Incentives: The Prod. Oper. Manag. 17(3): 267–284. Principal–Agent Model. Princeton University Press, Princeton, NJ. Menor, L. J., A. V. Roth, C. H. Mason. 2001. Agility in retail banking: Larson, R. C. 1987. Perspectives on queues: Social justice and the A numerical taxonomy of strategic service groups. Manuf. Serv. psychology of queueing. Oper. Res. 35(6): 895–905. Oper. Manage. 3(4): 273–292. Larson, R. C. 1990. The queue inference engine: Deducing queue Metters, R. D., F. X. Frei, V. A. Vargas. 1999. Measurement of statistics from transactional data. Manage. Sci. 36(5): 586–601. multiple sites in service firms with data envelopment analysis. Larson, R. C., E. J. Pinker. 2000. Staffing challenges in financial services. Prod. Oper. Manag. 8(3): 264–281. Melnick, E., P. Nayyar, M. Pinedo, S. Seshadri, eds. Creating Meuter, M. L., A. L. Ostrom, R. I. Roundtree, M. J. Bitner. 2000. Self- Value in Financial Services—Strategies. Operations and Technolo- service technologies: Understanding customer satisfacton with gies. Kluwer Academic Publishers, Norwell. MA, 327–356. technology-based service encounters. J. Mark. 64(7): 50–64. Lee, D. K. C., S. Lwi, K. F. Phoon. 2004. An equitable structure for Miller, M. H., O. Orr. 1966. A model of the demand for money by hedge fund incentive fees. J. Investing 13(3): 31–43. firms. Q. J. Econ. 80(3): 413–435. Lee, E.-J., D. B. Eastwood, J. Lee. 2004. A sample selection model of Muermann, A., U¨ .O¨ ktem. 2002. The near-miss management of consumer adoption of computer banking. J. Financ. Serv. Res. operational risk. J. Risk Finance 4(1): 25–36. 26(3): 263–275. Nair,S.K.,R.Anderson.2008.Aspecializedinventoryprobleminbanks: Levecq, H., B. W. Weber. 2002. Electronic trading systems: Strategic Optimizing retail sweeps. Prod. Oper. Manag. 17(3): 285–295. implications of market design choices. J. Organ. Comput. Electron. Commerce 12(1): 85–103. Nair, S. K., E. D. Hatzakis. 2010. Optimal management of sweeps to reduce sterile bank reserves. University of Connecticut Working Li, C. W., A. Tiwari. 2009. Incentive contracts in delegated portfolio paper. management. Rev. Financ. Stud. 22(11): 4681–4714. Neave, E. H. 1970. The stochastic cash balance problem with fixed costs Lynch, A. W., D. K. Musto. 1997. Understanding fee structures in for increases and decreases. Manage. Sci. 16(7): 472–490. the asset management business. NYU Working Paper No. FIN- 98-050, Department of Finance, Stern School of Business, New Netter, J. M., A. B. Poulsen. 2003. Operational risk in financial ser- York University. vice providers and the proposed Basel capital accord: An overview. Adv. Financ. Econ. 8: 147–171. Managed Funds Association. 2009. Sound practices for hedge fund managers. Available at http://www.managedfunds.org/files/ Nordgard, K. J., L. Falkenberg. 2010. Cost effectiveness in the asset pdf’s/MFA_Sound_Practices_2009.pdf (accessed date August management industry: An IT operations perspective. Pinedo, 3, 2009). M., ed. Operational Control in Asset Management—Processes and Costs. Palgrave Macmillan, New York, 132–150. Mandelbaum, A., A. Sakov, S. Zeltyn. 2001. Empirical analysis of a call center. Technical report. Available at http://iew3.technion. O’Brien, J. M. 2000. Estimating the value and interest rate risk of ac.il/serveng/References/ccdata.pdf (accessed date March 20, interest-bearing transaction deposits. FEDS Working Paper No. 2010). 00-53. Division of Research and Statistics, Board of Governors, Manski, C. F. 1995. Identification Problems in the Social Sciences. Har- Federal Reserve System. vard University Press, Cambridge, MA. O’Hara, M. 1998. Market Microstructure Theory. John Wiley and Sons, Massey, W. A., G. A. Parker, W. Whitt. 1996. Estimating the param- Malden, MA. eters of a nonhomogeneous Poisson process with linear rate. Oral, M., R. Yolalan. 1990. An empirical study on measuring oper- Telecommun. Syst. 5(2): 361–388. ating efficiency and profitability of bank branches. Eur. J. Oper. Massoud, N. 2005. How should central banks determine and control Res. 46(3): 282–294. their bank note inventory? J. Bank. Finance 29(12): 3099–3119. O¨ rmeci, E. L., O. Z. Aksin. 2010. Revenue management through Meester, G. A., A. Mehrotra, H. P. Natarajan, M. J. Seifert. 2010. dynamic cross-selling in call centers. Prod. Oper. Manag. 19(6): Optimal configuration of a service delivery network: An appli- 742–756. cation to a financial services provider. Prod. Oper. Manag. 19(6): Orr, D. 1974. A note on the uselessness of transactions demand 725–741. models. J. Finance 29(5): 1565–1572. Mehrotra, M., M. Dawande, V. Mookerjee, C. Sriskandarajah. 2010a Phillips, R. L. 2005. Pricing and Revenue Optimization.StanfordUni- Pricing and logistics decisions for a private-sector provider in versity Press, Stanford, CA. the cash supply chain. University of Texas at Dallas Working Phillips, R. L. 2010. Customized pricing. O¨ zer, O¨ ., R.L. Phillips, eds. paper. Handbook of Pricing Management. Oxford University Press, Mehrotra, M., M. Dawande., C. Sriskandarajah. 2010. A depository Oxford, U.K., to appear. institution’s optimal currency supply network under the Fed’s new guidelines: Operating policies, logistics and impact. Prod. Pinedo, M. L. 2009. Planning and Scheduling in Manufacturing and Oper. Manag. 19(6): 709–724. Services. 2nd edn. Springer, New York. Mehrotra, V., J. Fama. 2003. Call center simulation modeling: Meth- Pinedo, M. L. 2010. Global asset management: An introduction to its ods, challenges and opportunities. Proceedings of the 2003 Winter processes and costs. Pinedo, M. L. ed. Operational Control in Simulation Conference, 1: 135–143. Asset Management—Processes and Costs. Palgrave Macmillan, New York, 8–14. Mehrotra, V., O¨ .O¨ zlu¨ k, R. Saltzman. 2010. Intelligent procedures for intra-day updating of call center agent schedules. Prod. Oper. Pinedo, M. L., S. Seshadri, G. Shanthikumar. 2000. Call centers in Manag. 19(3): 353–367. financial services: Strategies, technologies and operations. Mel- Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society 663

nick, E. L., P. R. Nayyar, M. L. Pinedo, S. Seshadri, eds. Creating Smith, A. 1776. An Inquiry into the Nature and Causes of the Wealth of Value in Financial Services—Strategies, Operations and Technolo- Nations. Oxford University Press, Oxford, UK (2008 edition). gies. Kluwer Academic Publishers, Norwell, MA, 357–388. Smith, G. W. 1989. Transactions demand for money with a stochastic Rajamani, D., H. N. Geismar, C. Sriskandarajah. 2006. A framework time-varying interest rate. Rev. Econ. Stud. 56(4): 623–633. to analyze cash supply chains. Prod. Oper. Manag. 15(4): 544– Smith, J. S., K. R. Karwan, R. E. Markland. 2007. A note on the 552. growth of research in service operations management. Prod. Record Jr., E. E., M. A. Tynan. 1987. Incentive fees: The basic issues. Oper. Manag. 16(6): 780–790. Financ. Anal. J. 43(1): 39–43. So, K. C., C. Tang, R. Zavala. 2003. Models for improving Reed, J. 2009. The G/GI/N Queue in the Halfin-Whitt Regime. Ann. team productivity at the Federal Reserve Bank. Interfaces Appl. Probab. 19(6): 2211–2269. 33(2): 25–36. Ridley, A. D., W. Massey, M. Fu. 2004. Fluid approximation of a Soteriou, A., S. A. Zenios. 1999a. Operations, quality and profitabil- priority call center with time-varying arrivals. Telecommun. Rev. ity in the provision of banking services. Manage. Sci. 45(9): 15: 69–77. 1221–1238. Robbins, T. R., D. J. Medeiros, P. Dum. 2006. Evaluating arrival rate Soteriou, A., S. A. Zenios. 1999b. Using data envelopment analysis uncertainty in call centers. Proceedings of the 2006 Winter Sim- for costing bank products. Eur. J. Oper. Res. 114(2): 234–248. ulation Conference, 2180–2187. Soyer, R., M. M. Tarimcilar. 2008. Modeling and analysis of call Robbins, T. R., D. J. Medeiros, T. P. Harrison. 2007. Optimal cross center arrival data: A Bayesian approach. Manage. Sci. 54(2): training in call centers with uncertain arrivals and global 266–278. service level agreements. Working paper, Smeal College of Spohrer, J., P. P. Maglio. 2008. The emergence of service science: Business, Pennsylvania State University. Toward systematic service innovations to accelerate co-creation Robichek, A. A., D. Teichroew, J. M. Jones. 1965. Optimal short term of value. Prod. Oper. Manag. 17(3): 238–246. financing decision. Manage. Sci. 12(1): 1–36. Sprenkle, C. M. 1969. The uselessness of transactions demand models. Roth, A. V., W. E. Jackson, III. 1995. Strategic determinants of service J. Finance 24(5): 835–847. quality and performance: Evidence from the banking industry. Manage. Sci. 41(11): 1720–1733. Sprenkle, C. M. 1977. The uselessness of transaction demand models: Comment. J. Finance 32(1): 227–230. Roth, A. V., L. J. Menor. 2003. Insights into service operations man- agement: A research agenda. Prod. Oper. Manag. 12(2): 145–164. State Street. 2009. Outsourcing investment operations: Managing expense and supporting strategic growth. State Street Vision Rothschild, M., J. Stiglitz. 1976. Equilibrium in competitive insur- Series, May. Available at http://www.statestreet.com/ ance markets: An essay on the economics of imperfect knowledge/vision/2009/2009_oio.pdf (accessed date August information. Q. J. Econ. 90(4): 629–649. 22, 2009). Rudd, A., M. Schroeder. 1982. The calculation of minimum margin. Steckley, S. G., S. G. Henderson, V. Mehrotra. 2004. Service system Manage. Sci. 28(12): 1368–1379. planning in the presence of a random arrival rate. Working Sampson, S. E., C. M. Froehle. 2006. Foundations and implications paper, Department of Operations Research and Industrial of a proposed unified services theory. Prod. Oper. Manag. 15(2): Engineering, Cornell University, Ithaca, NY. 329–343. Steckley, S. G., S. G. Henderson, V. Mehrotra. 2009. Forecast errors in Schneider, A. 2010. Managing costs at investment management service systems. Probab. Eng. Inf. Sci. 23(2): 305–332. firms. Pinedo, M., ed. Operational Control in Asset Management Stiglitz, J. 1977. Monopoly, non-linear pricing and imperfect Processes and Costs. Palgrave Macmillan, New York, 42–58. information: The insurance market. Rev. Econ. Stud. 44(3): Seiford, L. M., J. Zhu. 1999. Profitability and marketability of the top 407–430. 55 U.S. commercial banks. Manage. Sci. 45(9): 1270–1288. Stiglitz, J. E., A. Weiss. 1981. Credit rationing in markets with im- Sethi, S. P., G. L. Thompson. 1970. Applications of mathe- perfect information. Am. Econ. Rev. 71(3): 393–410. matical control theory to finance: Modeling simple dynamic Stoll, H. R. 2006. Electronic trading in stock markets. J. Econ. cash balance problems. J. Financ. Quant. Anal. 5(4–5): 381–394. Perspect. 20(1): 153–174. Sheehan, R. G. 2004. Valuing core deposits. Working paper, Depart- Stoll, H. R. 2008. Future of securities markets: Competition or con- ment of Finance, University of Notre Dame. solidation? Financ. Anal. J. 64(6): 15–26. Shen, H., J. Z. Huang. 2005. Analysis of call centre arrival data using Stolletz, R. 2003. Performance Analysis and Optimization of Inbound singular value decomposition. Appl. Stoch. Models Bus. Ind. Call Centers. Lecture Notes in Economics and Mathematical Systems 21(3): 251–263. 528. Springer Verlag, Berlin. Shen, H., J. Z. Huang. 2008a. Interday forecasting and intraday up- dating of call center arrivals. Manuf. Serv. Oper. Manage. 10(3): Straub, D. W., R. J. Welke. 1998. Coping with systems risk security 391–410. planning models for management decision making. MIS Q. 22(4): 441–470. Shen, H., J. Z. Huang. 2008b. Forecasting time series of inhomoge- neous Poisson processes with application to call center Stulz, R. M. 2007. Hedge funds: Past, present, and future. J. Econ. workforce management. Ann. Appl. Stat. 2(2): 601–623. Perspect. 21(2): 175–194. Sherman, H. D., F. Gold. 1985. Bank branch operating efficiency: Talluri, K., G. Van Ryzin. 2004. Revenue management under a gen- Evaluation with data envelopment analysis. J. Bank. Finance eral discrete choice model of consumer behavior. Manage. Sci. 9(2): 297–315. 50(1): 15–33. Sherman, H. D., G. Ladino. 1995. Managing bank productivity using Talluri, K. T., G. J. Van Ryzin. 2005. The Theory and Practice of Revenue data envelopment analysis (DEA). Interfaces 25(2): 60–73. Management. Springer, New York. Shostack, G. L. 1984. Designing services that deliver. Harv. Bus. Rev. Taylor, J. W. 2008. A comparison of univariate time series methods 62(1): 133–139. for forecasting intraday arrivals at a call center. Manage. Sci. Slepicka, K.L., G.A. Sporer. 1981. Application of simulation to the 54(2): 253–265. banking industry. Proceedings of the 1981 Winter Simulation Con- Thomas, L. C. 2009. Consumer Credit Models: Pricing, Profit and Port- ference, 481–485. folios. Oxford University Press, Oxford, UK. Hatzakis, Nair, and Pinedo: Operations in Financial Services—An Overview 664 Production and Operations Management 19(6), pp. 633–664, r 2010 Production and Operations Management Society

Thomas, L. C., D. B. Edelman, J. N. Cook. 2002. Credit scoring and Weber, B. W. 2006. Adoption of electronic trading at the Interna- its applications. Monographs on Mathematical Modeling & tional Securities Exchange. Decis. Support Syst. 41(4): 728–746. Computation. SIAM, Philadelphia. Wei, R. 2006. Quantification of operational losses using firm-specific Thomke, S. 2003. R&D comes to services: Bank of America’s path- information and external database. J. Oper. Risk 1(4): 1–33. breaking experiments. Harv. Bus. Rev. 81(4): 70–79. Weinberg, J., L. D. Brown, J. R. Stroud. 2007. Bayesian forecasting of Thompson, G. 1993. Accounting for the multi-period impact of an inhomogeneous Poisson process with applications to call service when determining employee requirements for labor center data. J. Am. Stat. Assoc. 102(480): 1185–1198. scheduling. J. Oper. Manage. 11(3): 269–287. Weitzman, M. 1968. A model of the demand for money by firms: Thompson, G. M. 1998. Labor scheduling. Part 1: Forecasting de- Comment. Q. J. Econ. 82(1): 161–164. mand. Cornell Hotel Rest. Ad. 39(5): 22–31. Whalen, E. L. 1966. A rationalization of the precautionary demand Tirole, J. 1988. The Theory of Industrial Organization. The MIT Press, for cash. Q. J. Econ. 80(2): 314–324. Cambridge, MA. Whistler, W. D. 1967. A stochastic inventory model for rented Tobin, J. 1956. The interest elasticity of transactions demand for equipment. Manage. Sci. 13(9): 640–647. cash. Rev. Econ. Stat. 38(3): 241–247. Whitman, M. E. 2004. In defense of the realm: Understanding threats to information security. Int. J. Inf. Manage. 24(1): 43–57. Trench, M. S., S. P. Pederson, E. T. Lau, L. Ma, H. Wang, S. K. Nair. 2003. Managing credit lines and prices for Bank One credit Whitt, W. 1999a. Dynamic staffing in a telephone call center aiming cards. Interfaces 33(5): 4–21. to immediately answer all calls. Oper. Res. Lett. 24(5): 205–212. Tych, W., D. J. Pedregal, P. C. Young, J. Davies. 2002. An unobserved Whitt, W. 1999b. Predicting queueing delays. Manage. Sci. 45(6): component model for multi-rate forecasting of telephone call 870–888. demand: The design of a forecasting support system. Int. J. Whitt, W. 2006. Fluid models for many-server queues with aban- Forecast. 18(4): 673–695. donments. Oper. Res. 54(1): 37–54. Vassiloglou, M., D. Giokas. 1990. A study of the relative efficiency of Wuebker, G., M. Koderisch, D. Smith-Gallas, J. Baumgarten. 2008. bank branches: An application of data envelopment analysis. Price Management in Financial Services: Smart Strategies for J. Oper. Res. Soc. 41(7): 591–597. Growth. Gower Publishing, Farnham, UK. Vial, J.-P. 1972. A continuous time model for the cash balance prob- Xu, W. 2000. Long range planning for call centers at Fedex. J. Bus. lem. Szego, G. P., K. Shell, eds. Mathematical Models in Investment Forecast. Methods Syst. 18(4): 7–11. and Finance. North-Holland, Amsterdam, The Netherlands, Xue, M., L. M. Hitt, P. T. Harker. 2007. Customer efficiency, channel 244–291. usage, and firm performance in retail banking. Manuf. Serv. Vickson, R. G. 1985. Simple optimal policy for cash management: Oper. Manage. 9(4): 535–558. The average balance requirement case. J. Financ. Quant. Anal. Zenios, C. V., S. A. Zenios, K. Agathocleous, A. C. Soteriou. 1999. 20(3): 353–369. Benchmarks of the efficiency of bank branches. Interfaces 29(3): Voss, C., A. V. Roth, R. B. Chase. 2008. Experience, service opera- 37–51. tions strategy, and services as destinations: Foundations Zhu, J. 2003. Imprecise data envelopment analysis (IDEA): A review and exploratory investigation. Prod. Oper. Manag. 17(3): and improvement with an application. Eur. J. Oper. Res. 144(3): 247–266. 513–529. Weber, B. W. 1999. Next-generation trading in futures markets: A Zohar, E., A. Mandelbaum, N. Shimkin. 2002. Adaptive behavior of comparison of open outcry and order matching systems. impatient customers in tele-queues: Theory and empirical sup- J. Manage. Inf. Syst. 16(2): 29–45. port. Manage. Sci. 48(4): 566–583.