Demand allocation in systems with multiple inventory locations and multiple demand sources

Saif Benjaafar, Yanzhi Li, Dongsheng Xu, Samir Elhedhli

Research output: Contribution to journalArticlepeer-review

17 Scopus citations


We consider the problem of allocating demand that originates from multiple sources among multiple inventory locations. Demand from each source arrives dynamically according to an independent Poisson process. The cost of fulfilling each order depends on both the source of the order and its fulfillment location. Inventory at all locations is replenished from a shared production facility with a finite production capacity and stochastic production times. Consequently, supply lead times are load dependent and affected by congestion at the production facility. Our objective is to determine an optimal demand allocation and optimal inventory levels at each location so that the sum of transportation, inventory, and backorder costs is minimized. We formulate the problem as a nonlinear optimization problem and characterize the structure of the optimal allocation policy. We show that the optimal demand allocations are always discrete, with demand from each source always fulfilled entirely from a single inventory location. We use this discreteness property to reformulate the problems as a mixed-integer linear program and provide an exact solution procedure. We show that this discreteness property extends to systems with other forms of supply processes. However, we also show that supply systems exist for which the property does not hold. Using numerical results, we examine the impact of different parameters and provide some managerial insights.

Original languageEnglish (US)
Pages (from-to)43-60
Number of pages18
JournalManufacturing and Service Operations Management
Issue number1
StatePublished - Dec 2008


  • Facility location
  • Make-to-stock queues
  • Optimal demand allocation
  • Production-inventory systems


Dive into the research topics of 'Demand allocation in systems with multiple inventory locations and multiple demand sources'. Together they form a unique fingerprint.

Cite this