The impact of operations anticipations on the quality of stochastic location-allocation models

It is now a common practice to use optimization models, such as location-allocation models, to support the design of supply chain networks (SCN). The value creation potential of a SCN design must be evaluated in terms of capital expenditures, but also of the operating revenues and expenses incurred during the planning horizon considered. The design model used should therefore be formulated to anticipate these revenues and expenses (relatively) accurately over the planning horizon. In classical location-allocation models, the aggregate flow and throughput variables used yield very crude anticipations. It was never shown that they lead to the best SCN design that one should expect. This paper draws on the stochastic multi-period location-transportation problem (SMLTP) for studying the impact of various types of operations anticipations on the quality of the SCN designs obtained. Since accurate anticipations yield more complex models, solvability is also an issue. Several alternative SCN design models based on more detailed anticipations than the ones embedded in classical location-allocation models are proposed and tested. Accuracy-solvability trade-offs are explored and recommendations are made on the modeling strategy to use to get better SCN designs.