APPLICATION OF DEMAND FORECASTING MODELS TO INVENTORY MANAGEMENT OPTIMIZATION: A DATA-DRIVEN APPROACH
TL;DR
The evidence supports an integrated forecasting–inventory perspective in which predictive performance and inventory policy are evaluated jointly, and this approach can improve the balance between product availability, working capital, and logistics costs, although results remain dependent on demand characteristics, model calibration, and the operating context.
Abstract
Demand forecasting is a central input to inventory planning because replenishment decisions depend not only on expected demand but also on the uncertainty surrounding that estimate. This narrative review examines how data-driven forecasting models can contribute to reducing excess inventory and stockouts while maintaining adequate service levels. The literature reviewed indicates that ensemble learning, machine-learning models, probabilistic approaches, and hybrid forecasting methods can improve demand estimation, particularly when seasonality, promotions, economic indicators, and other exogenous variables are incorporated. However, greater statistical accuracy does not automatically translate into lower inventory costs or higher service levels. Operational performance depends on linking forecast outputs and residual uncertainty to safety-stock calculations, reorder points, order-up-to policies, and other replenishment rules. The evidence therefore supports an integrated forecasting–inventory perspective in which predictive performance and inventory policy are evaluated jointly. This approach can improve the balance between product availability, working capital, and logistics costs, although results remain dependent on demand characteristics, model calibration, and the operating context.