Forecasting Demand for Optimal Inventory with Long Lead Times: An Automotive Aftermarket Case Study

Accuracy in predicting customer demand is essential to building an economic inventory policy under periodic review, long lead-time, and a target fill rate. This study uses inventory and customer service level as a stock control metric to evaluate the forecast accuracy of different simple to more complex predictive analytical techniques. We show how traditional forecast error measures are inappropriate for inventory control, despite their consistent usage in many studies, by evaluating demand forecast performance dynamically with customer service level as a stock control metric that includes inventory holdings costs, stock out costs, and fill rate service levels. A second contribution includes evaluating the utility of introducing more complexity into the forecasting process for an automotive aftermarket parts manufacturer and the superior inventory control results using the Prais-Winsten, an econometric method, for non-intermittent demand forecasting with long-lead times. This study will add to the limited case study research on demand forecasting under long lead times using stock control metrics, dynamic model updating, and the Prais-Winsten method for inventory control.