USE OF SIMULATION TO DETERMINE CASHIER STAFFING POLICY AT A RETAIL CHECKOUT Edward J. Williams Mohamed Karaki Craig Lammers Industrial and Manufacturing Systems Engineering 2330 Engineering Complex University of Michigan – Dearborn 4901 Evergreen Road Dearborn, Michigan 48128 U.S.A. williame@umdsun2.umd.umich.edu ABSTRACT Both queuing theory analysis and discrete-event process simulation have often been used, sometimes jointly, to analyze and improve the performance of queuing systems. Queuing theory provides closed-form solutions for various canonical queuing configurations, whereas discrete-event process simulation is highly valuable for analysis of many queuing systems beyond the reach of such closed-form solutions. Since queues are extremely ubiquitous, both queuing theory analysis and discrete-process simulation are frequently and beneficially used by analysts, engineers, and business managers. During the study presented in this paper, discrete-event process simulation was used to analyze, specify, and improve operational policies in a large retail store. Results of the model guided store management toward policies, ultimately proved successful in practice, governing the thresholds of congestion warranting the opening and closing of cash-register lanes during a retail- business day. INTRODUCTION AND OUTLINE The objective of this study, as specified jointly by the simulation analysts and a retail store manager, was to assess staffing policies at checkout counters, and thence to select a policy balancing staffing costs against the costs (e.g., in goodwill and repeat business) of overly long customer delays within the checkout queues. Since traditional, operations-research-based queuing theory methods provided only approximations to the non-canonical queuing configuration within the retail store, the study relied extensively upon the power of discrete-event process simulation to provide accurate assessments and comparisons among candidate staffing and operational policies. Applications of discrete-process simulation to retail stores seem rare in the literature, although the issues involved are analogous to those arising in analysis of a customer service center receiving orders by mail, facsimile, and telephone (Chin and Sprecher 1990) or analysis of a fast-food restaurant (Farahmand and Martinez 1996). First, this paper provides an overview of the queuing system in the context of retail store management considerations and concerns. Next, we describe the input data collection required in support of the study, followed by the construction, verification, and validation of the simulation model. We then conclude by describing the results of the study, the response of the client to receipt of these results, and likely directions for future extensions of this work. OVERVIEW OF THE QUEUING SYSTEM IN CONTEXT In the United States culture, household pets (most often dogs or cats) are very popular, and often achieve a psychological status amounting nearly to “members of the family.” (Beck and Katcher 1996). Research purporting to prove that ownership of a pet enhances longevity, especially among the elderly and/or infirm (Elgin 1990), has earned much publicity in North America. For several generations, local entrepreneurs turned shopkeepers have operated “pet stores” which sell the actual pet (dog, cat, goldfish, hamster,….) and also sell pet supplies (pet food, leashes, feeding dishes, beds and coverlets,….). Perhaps inevitably, large chain stores have eagerly seized this retailing niche opportunity, just as chain stores have moved into niches such as groceries, hardware, books, and home- office supplies (Davids 1997). The original “Pet Supplies ‘Plus’” store, opened in 1988 by Jack Berry and Harry Shallop, has expanded into a chain of more than 175 stores dispersed among 19 of the 50 states (Berman 1998). Like any retail store, the local Pet Supplies “Plus” store analyzed in this study must devote considerable managerial attention to checkout operations – that is, the staffing level at the checkout registers necessary to provide acceptable promptness of service to customers without undue overhead expense. Long checkout times, and high variance (hence unpredictability) of checkout times annoy customers; annoyed customers, in turn, represent both potential loss of future sales and bad publicity for the store. Psychologically, customers are more concerned with the time spent waiting in line for checkout service (w Q in traditional queuing-theory notation) than with total wait time plus checkout-service time (w in traditional queuing- theory notation). Yet attempting to reduce w Q to zero (reducing w to zero is manifestly impossible) would Proceedings 14th European Simulation Symposium A. Verbraeck, W. Krug, eds. (c) SCS Europe BVBA, 2002