Evaluation of forecast models used for inventory control during a product’s life cycle: A simulation study

This article analyses the consequences of using different exponential smoothing models in inventory control. The evaluation of forecast models has received much attention in the literature. In most articles, forecast models have been evaluated using statistical criteria, such as average forecast error and mean square error. From the point of view of inventory control, however, it is more interesting to examine the forecast model’s impact on inventory control performance as measured by, for example, the resulting costs-including holding costs, order costs and shortage costs. The method used is stochastic simulation, with which it is possible to simulate the practical application of forecast models together with theoretical inventory models. The simulations are carried out for a product whose probability distribution of demand is the normal distribution with (a) a mean described by a traditional life cycle, to which both 6- and 12-month seasons have been added, and (b) both constant and time-dependent standard deviations. In this analysis, it is assumed that inventory is controlled by an (s,Q) system, where the order quantity is calculated using the EOQ formula, and the reorder point is determined using an inventory model based on the minimisation of order costs, holding costs and shortage costs.