When Does Pooling Pay? Credibility and Resolution under Forgetting in Intermittent-Demand Forecasting
The paper introduces TSB-HB, a hierarchical Bayesian extension of the Teunter–Syntetos–Babai (TSB) method for intermittent demand forecasting. It uses hierarchical priors to allow partial pooling across items, stabilizing estimates for sparse or cold-start series while maintaining item heterogeneity. This approach significantly improves forecast accuracy and distributional reliability compared to standard TSB.