Setting accuracy targets for short-term judgemental sales forecasting

Traditionally, the quality of a forecasting model is judged by how it compares, in terms of accuracy, to alternative models. However, by providing a relative measure, no indication is given as to how much scope there might be for improvements beyond the benchmark model. When judgemental methods are used alongside simple forecasting models, the scope for such improvements is considerable and difficult to benchmark. Derivation of targets for forecasting quality is thus not straightforward. The approach taken in this paper is to consider forecast error as consisting of irreducible error due to intrinsic unpredictable uncertainty, and error due to less than perfect modelling, estimation and forecasting. As the intrinsic uncertainty presents a bound on forecast accuracy, our derivation of an accuracy target is based on the measurement of this irreducible uncertainty. The motivation and data for this case-study were taken from the short-term sales forecasting process of a major, international high-technology manufacturer.