Seasonal adjustment of inventory demand series: A case study

This paper analyzes procedures for seasonal adjustment of inventory demand series at a large US auto parts distributor, BPX Holding Corporation of Houston, TX. The company's forecasting system made no attempt to classify demand series as seasonal or nonseasonal. All demand series were assumed to be seasonal. They were seasonally-adjusted using a multiplicative decomposition procedure, then forecast with exponential smoothing. We show that simple methods of identifying seasonal series, coupled with an additive decomposition procedure, can make significant reductions in forecast errors and safety stock investment. We also discuss forecasting implementation problems in inventory control systems.