Use of cost and accuracy measures in forecasting method selection: A physical distribution example

A difficult inventory control decision is that encountered by electronics warehouses which must stock a large variety of items in a rapidly changing, high cost industry. Electronic component distributors face high variability in demand, a lack of sufficient historical data and short product life cycles. The selection of accurate forecasting techniques for such a data set is of major importance. Four forecasting approaches were tested. They include single exponential smoothing with different smoothing constants as well as double exponential smoothing, adaptive response rate exponential smoothing and median of the historical data as the forecast. The selection process requires the utilization of an appropriate metric. In this study, in addition to the traditional accuracy metric an added criterion, the use of an asymmetric economic measure of forecasting accuracy is proposed and tested. The forecasting methods utilized accuracy and economic measures, with a large data set. The different measures yielded different answers. The economic measure can provide managers with insights not directly available from some of the traditionally used forecasting error metrics for this decision.