Estimation in supply chain inventory management

Differences in estimation or forecasting procedures could produce dramatically different parameter estimates in supply chain inventory management. We show, for example, that determining when to introduce estimates of lead times in the calculation of the variance of demand during lead time can yield dramatically different safety stocks and order-up-to levels. Also, calculations of supply chain variance amplification using a firewalled, sequential chain execution differ markedly from an analysis that considers a k-echelon analysis as a whole, k⩾2. There is also the issue of forecasting lumpy demand when negative orders are not allowed. Our research compares the results in the recent literature and shows how apparently equivalent estimation procedures concerning demand during lead time (for example, using separate historical lead time and demand rate data versus directly using historical data of demand during lead time) are not equivalent; also, that the conventional exponential smoothing forecasting may not be appropriate at the higher echelons of supply chains where lumpy demand frequently occurs.