This paper addresses the problem of storage assignment in a warehouse characterized by multi-command picking and served by milkrun logistics. In such a logistic system, vehicles circulate between the warehouse and the production facilities of the plant according to a pre-defined schedule, often with multiple cycles (routes) serving different departments. We assume that a request probability can be assigned to each item and each cycle, which leads to a special case of the correlated storage assignment problem. A MIP model is proposed for finding a class-based storage policy that minimizes the order cycle time, the average picking effort, or a linear combination of these two criteria. Computational experiments show that our approach can achieve an up to 36–38% improvement in either criterion compared to the classical COI-based strategy.