Variance laws for inventory management

In forecasting for inventory management, the demand variance for each stock-keeping unit must be estimated. Direct estimation of variance from the data is problematic and many systems use a relationship linking the variance of demand to its mean. In this paper, a theory of such functional relationships is developed for compound Poisson demand. The assumptions behind the compound Poisson model are examined and the robustness of functional relationships to departures from those assumptions are analysed. The paper concludes with a discussion of the relevance of the results to demand forecasting and inventory centralisation modelling.