Demand Distortions and Capacity Allocation Policies

Abstract

In this paper, we describe a study of the consequences of capacity allocation mechanisms at a large semiconductor manufacturer. We use a panel data set with 300,000 observations spanning five years of forecasting and order transactions between a supplier and customers at the supplier's manufacturing facility. Using Mixed Linear Models (MLM), we study the mutual interplay between a supplier's capacity allocation mechanism and customers' demand forecasts and orders. The results of our model suggest that the interaction causes two types of distortions: inter-temporal and cross-sectional. Temporal forecast distortions result in forecast churn (short-term volatility in forecasts), undesirable forecast smoothing (a.k.a., batching), and customers exiting the facility. Cross-sectional forecast distortion (contagion) is characterized by temporal forecast distortions spreading across individual customers' forecasts, attributable to the negative externalities imposed by commonly used capacity allocation policies. We present empirical tests for the presence and significance of these distortions in our data and note the impact of churn on buffer inventory stocks and contagion's effect on risk pooling efforts. Our findings have implications for contracting, capacity planning, and CPFR initiatives in supply chains.