Evaluation of Variance-Covariance-Matrix Application for Identifying Temporary Demand Shifts in a Forecast Evolution System.

Abstract

Customer provided demand forecasts are in many production systems the basis for production planning and significantly influence the performance of the respective system. To improve the forecast accuracy, forecast correction methods can be applied to mitigate the effects of systematic customer behavior when providing their forecasts. For applying such forecast correction methods, it is important to identify different kinds of systematic customer behavior in generating their forecasts. In this paper we introduce a Variance-Covariance-Matrix analysis, which is based on forecast evolution literature, to identify demand shifting between periods. On the one hand we derive analytically the systematic effect of demand shifts within forecasts on parts of the Variance-Covariance-Matrix. On the other hand, we provide a Monte Carlo simulation to identify the stochastic disturbance effects on the Variance-Covariance-Matrix if stochastic forecast updates are realized. The numerical study shows that this analysis is a promising tool to identify such demand shifting between periods, however, there is still a lot of further research necessary.