Who Directs the Control Tower? Customer vs. Supplier-Managed Inventory in Supply Chains

Information distortion and incentive misalignment often undermine supply chain performance. In response, many firms adopt a “control tower” approach, whereby one party assumes responsibility for replenishment decisions across the supply chain. Industry examples showed that this role may be assigned to the customer, Customer-Managed Inventory (CMI), or to the supplier, Supplier-Managed Inventory (SMI). This paper investigates when one governance structure is preferable. We study a two-echelon, one-to-many supply chain consisting of a single supplier and a customer operating multiple warehouses. We characterize optimal base-stock policies and equilibrium holding-cost-sharing parameters under CMI and SMI and benchmark them against decentralized operations and a fully centralized system. Two primary forces shape the choice between CMI and SMI: the relative shortage penalty costs of the two parties and the structure of the customer’s distribution network. Specifically, CMI is preferred when the customer faces relatively higher shortage penalties and operates in a more asymmetric environment—for example, when demand is concentrated in a subset of its warehouses or when replenishment lead times vary substantially across these locations. In contrast, SMI is favored under opposite conditions. We further examine how intertemporal and interspatial demand correlations affect this comparison. A key finding is that, although neither CMI nor SMI achieves the first-best outcome, their efficiency loss relative to full centralization is typically small. Calibrated numerical studies based on four industry cases disentangle the incremental value of information sharing, decision-right transfer, and cost sharing, showing that most performance gains stem from decision-right transfer supported by appropriate holding-cost sharing.

This paper was accepted by Jayashankar Swaminathan, operations management.

Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2025.02137.