How to use aggregation and combined forecasting to improve seasonal demand forecasts

Standard forecasting methods that are designed to cope with seasonal demand often are no longer applicable in practice. Due to growing assortments and shorter product life cycles, demand data may show too high variation or may be insufficient to construct reliable forecast models at the individual item level. In this article we present alternative forecasting methods that are based on using demand information from a higher aggregation level and on combining forecasts. Sales data from two prominent Dutch wholesalers are used to illustrate the drawbacks of the standard seasonal forecasting methods and to demonstrate the potential of the new methods. The average reduction in forecast error (in terms of MSE) turns out to be three times as large as reported in earlier studies on common seasonal patterns.