ABSTRACT
Tracking methods are used in short-term forecasting, where one is trying to reduce to the minimum the need for profound understanding of underlying processes, and where ease of computation is an important consideration. Analysis of forecast errors indicated that participants were generally fairly good. Forecasting systems employing exponential smoothing usually have a value of 0.2 or less for the smoothing constant a. However, a problem arises if the system encounters a sudden change in the underlying circumstances. The practical difficulties are to reduce the arithmetical effort involved in the more sophisticated techniques, and the ultimate aim is to produce a robust predictor, preferably similar to the non-seasonal predictors. A forecast made with no associated measure of its reliability is incomplete and may at times be misleading. In production control, one basic function of buffer stocks is to absorb anticipated fluctuations in demand or departure from forecast.
