The Use of Cumulative Sum (Cusum) Techniques for the Control of Routine Forecasts of Product Demand

Abstract

The cumulative sum technique has proved its value in process control and this paper discusses its use when applied to sales forecasting systems. The two main applications of cusum are (i) to detect when the pattern of customer demand changes, and (ii) to detect when a forecasting system is no longer adequate. The former use generally applies when demand is reasonably steady and the latter when changes are more frequent. The question of re-estimating the parameters used in the forecasting system is considered and, in particular, the problem of differentiating between a change in trend and a change in the mean level of demand is studied. Finally, it is shown how, the cusum control scheme can be simplified so that its operation is suitable for a computer.