Rolling horizon modeling vs. multi-period modeling

Rolling horizon modeling replaces multi-period safety stock optimization for planning and operational business cases. In multi-period safety stock optimization, each period is independent from the others. As a result, safety stock placement decisions can be different from one period to the next which can cause operational problems. In addition, each period is run as a separate SSO problem, increasing the solve time. Multi-period models can be useful when modeling seasonality, where you can treat safety stock placement for each season as independent.

With rolling horizon modeling, you work with a single period model. The placement decisions made in the historic horizon are passed to the planning horizon. In the planning horizon, the inventory policies are fixed. It is only the policy parameters that change. This approach provides a more practical solution when applied to operational business practices. Rolling horizon modeling also handles time shifts better, especially when you have demand and forecast data based on shorter forecast time buckets, such as weekly aggregation.