PRODUCTION AND OPERATIONS MANAGEMENT Vol. 8, No. I, Spring 1999 Printed in U.S.A. A COMPARISON OF PRODUCTION SCHEDULING POLICIES ON COSTS, SERVICE LEVEL, AND SCHEDULE CHANGES * RICHARD METTERS AND VICENTE VARGAS Cox School of Business, Southern Methodist University, Dallas, Texas 152750333, USA Goizueta Business School,Emory University, Atlanta, Georgia 30322-27 10, USA We consider a single product, single level, stochastic master production scheduling (MPS) model where decisions are made under rolling planning horizons. Outcomes of interest are cost, service level, and schedule stability. The subject of this research is the MPS control system: the method used in determining the amount of stock planned for production in each time period. Typically, MPS control systems utilize a single buffer stock. Here, two MPS dual-buffer stock systems are developed and tested by simulation. We extend the data envelopment analysis ( DEA) methodology to aid in the evaluation of the simulation results, where DEA serves to increase the scope of the experimental design. Results indicate that the dual-buffer control systems outperform existing policies. (MASTER PRODUCTION SCHEDULING; DATA ENVELOPMENT ANALYSIS; PERI- ODIC REVIEW INVENTORY THEORY; SIMULATION) 1. Introduction We consider a single product, single-level master production scheduling ( MPS) model with no capacity restrictions. Demand is stochastic and time-varying. Production decisions are made under rolling planning horizons. A make-to-stock environment is presumed. System performance is measured by three criteria: the cost of inventory holding and production setup, customer service level (fill rate), and changes in the production schedule (schedule instability). The inclusion of schedule instability as a performance metric bears further comment. Schedule instability was recognized as a significant practical problem for material re- quirements planning ( MRP) systems early in the development of MRP (Steele 1975). MRP systems utilize an MPS to translate independent demand for end items to dependent demand for components and raw materials. Because of the lead times and scheduling complexities of MRP systems, production in an MPS must be planned for future periods. As new infor- mation concerning demand is gathered, the planned future production amounts may change. Changes to the MPS, while desirable in light of updated demand information, may cause a cascade of changes as requirements are exploded down through the bill-of-ma- * Received March 1997; revision received October 1997; accepted March 1998. 76 1059-1478/99/0801/076$1.25 Copyright 8 1999, Pmduction and Operations Management Society