Process Location and Product Distribution with Uncertain Yields

We present a framework to analyze the process location and product distribution problem with uncertain yields for a large multinational food processing company. This problem consists of selecting the location of processes, the assignment of products, and the distribution of production quantities to markets in order to minimize total expected costs. It differs from the traditional facility location problem due to characteristics that are inherent to process industry sectors. These include significant economies of scale at high volumes, large switchover times, and production yield uncertainty. We model the problem as a nonlinear mixed-integer program. A challenging aspect of this problem is that the objective function is neither convex nor concave. We develop an exact approach to linearize the objective function. We present heuristics to solve the problem and also construct lower bounds based on a reduction of the constraint set to evaluate the quality of the solutions. This framework has been used to make process choice and product allocation decisions at the food processing company, and the estimated annual cost savings are around 10%, or $50 million. In addition, the insights from the model have had a significant strategic and organizational impact at this company. Our framework and conclusions are relevant to other industrial sectors with similar characteristics, such as pharmaceuticals and specialty chemical manufacturers.