Predicting the cost performance of inventory control systems by retrospective simulation

When an inventory-control system is designed to utilized statistical demand information in computing an (s,S) policy, it may be necessary to use performance forecasts to calibrate the relationship between control parameters and performance measures. Using a simulation model to test prediction by retrospective simulation we estimated the bias and variance of forecasts for expected costs and other operating characteristics. The results show the sensitivity of the prediction bias to changes in demand, cost, and lead time parameters and to choices made in system design.