Proceedings of the 1998 Winter Simulation Conference D.J. Medeiros, E.F. Watson, J.S. Carson and M.S. Manivannan, eds. DISCRETE TIME SIMULATION OF AN EQUIPMENT RENTAL BUSINESS R. Alan Bowman Graduate Management Institute Union College Schenectady, NY 12308, U.S.A. Ira J. Haimowitz Pfizer, Inc. 235 East 42nd Street, 7th floor New York, NY, 10017, U.S.A Robert M. Mattheyses A. Yonca ¨ Ozge Mary C. Phillips GE Corporate R & D One Research Circle Niskayuna, NY, 12309, U.S.A. ABSTRACT In this work we report the results of a discrete time simulation model that we developed for an equipment rental business to study the impact of business decisions. The whole tool consisted of a user interface that enabled efficient viewing and modifying of the input data, executing the simulation program, and viewing the output reports. Utilizing this we applied cost/benefit analysis to the results of the simulation runs and identified profitable investment alternatives for the business. We also measured their asset population in terms of their profitability and quantified the relation between utilization, repair times, and responsiveness to the customers. 1 INTRODUCTION In this paper we report a simulation study of the dependence of various performance measures in an equipment rental business on controllable factors such as inventory levels and repair times. The size of the problem (with thousands of asset types and ten thousands of individual assets) and the complexity of the relation between the performance measures and controllable factors (decision variables) led us to use simulation in our analysis. This is a widely used technique to analyze such complex systems; using a finite but sufficiently long simulation run one can closely estimate performance at any setting of the decision variables. This approach when combined with gradient estimation (Ho and Cao 1991, Rubinstein and Shapiro 1993) becomes more powerful and can be used in the solution of stochastic optimization (Robinson 1996, Rubinstein and Shapiro 1993) and equilibrium problems (G¨ urkan et al. 1998). Due to time and budget limitations commonly present in many real-world projects and the additional effort that would have been necessary to carefully optimize the system under study, we could not apply a simulation optimization approach. Instead, we applied Cost/Benefit analysis (Grant et al. 1990) to the results of the simulation runs and determined alternative operating policies that substantially improved performance, in this case utilization of assets (fraction of assets being rented at any given time), responsiveness to customers (i.e., fill rate), and return on investment. From an aggregate point of view there is a trade-off between utilization and responsiveness to customers. The leaders of the business tended to focus on utilization perhaps at the expense of investing to meet customer demand. In this study we showed that a careful choice (by quantifying the contribution to lost revenue of various asset types and using this information in the financial analysis) of asset types to invest in improved both measures. The remainder of the paper is organized as follows: in Section 2 we motivate the work by describing the business and the problems they were facing. In analyzing the business we used a special simulation technique: discrete time simulation. In Section 3 we discuss the reasons behind choosing discrete time simulation. Section 4 describes interesting features of the problem resulting from unavailability of perfect data, bill of material structure of the orders, and efficiency considerations. We also describe how we handled these difficulties. In Section 1505