A hierarchical forecasting approach for pricing decisions in capacity-constrained cold storage supply chains

Accurate forecasting plays a pivotal role in supply chain management (SCM), especially within capacity-constrained cold storage systems where pricing, procurement, production, and distribution decisions are highly interdependent. In such contexts, warehouse occupancy both affects and is affected by upstream supply processes, such as replenishment and production planning, as well as downstream operations, including order fulfillment and distribution flows. This study develops a hierarchical forecasting framework combining behavioral time-series profiling, profile-based model comparison, forecast reconciliation, and decision-oriented evaluation. The framework is applied to six years of weekly stock data covering 70 products grouped into nine operational families. Candidate models are evaluated through rolling-origin out-of-sample validation, and family-level forecasts are reconciled with total warehouse occupancy using alternative hierarchical methods. The results show that forecasting performance varies across behavioral profiles and that MinT diagonal provides the lowest observed aggregate error measures, although bootstrap intervals indicate overlapping performance across reconciliation methods. Statistical rankings also do not fully coincide with tariff-oriented and scenario-based economic outcomes. These findings show that occupancy forecasts should be evaluated jointly in terms of accuracy, hierarchical coherence, uncertainty, and operational consequences. By integrating behavioral segmentation, hierarchical forecasting, and decision-oriented evaluation, this study demonstrates the importance of assessing forecasting systems based on their impact on interconnected supply chain decisions rather than predictive accuracy alone. The framework provides both methodological contributions and managerial insights for managing capacity-constrained Cold storage supply chain.