Forecasting of lead-time demand variance: Implications for safety stock calculations

Lead-time demand forecasting constitutes the backbone of inventory control. Although there has been a considerable amount research on mean lead-time demand, less centered around variance, especially in case stochastic lead-times. This represents an important gap literature, given that safety stock calculations rely explicitly variance (or equivalently forecast error for unbiased estimators). We bridge this by exploring viability three strategies to estimate under lead-times: (1) aggregating per period errors over lead-time, which is classical approach; (2) considering aggregated (over lead-time) error; (3) resulting from temporally length) demand. Analytical results are derived first order autoregressive moving average ARMA(1,1) process both single exponential smoothing and minimum squared method. A numerical investigation assesses effects autocorrelation variability accuracy each strategy, conditions one outperforms others. The show strategy presented textbooks appears be least accurate one, except cases with high negative autocorrelation. An analysis control performance also reveals often leads higher costs lower service levels positive