Demand forecasting and cost performance in a model of a real manufacturing unit

Abstract

A series of experiments is reported in which the demand forecasting method used in a simulation model of an actual make-for-stock shop was varied, and the resulting impact on a cost function observed. The forecasting models employed were all of the multiple exponential smoothing type. Reducing the smoothing constant, i.e. changing the filtering characteristics, was found to lead to statistically significant cost savings across the full range of experimental conditions examined. However, the effect of increasing the order of the forecasting model, i.e. changing the assumption about the nature of the demand time series, was found to be sensitive to the stock-out cost rate used, with both increases and decreases in cost occurring.