SKU decomposition algorithms for the tactical planning in the FMCG Industry

Tactical planning MILP models for fast moving consumer goods companies are often intractable for realistic applications containing thousands of Stock-Keeping Units (SKUs). Decomposition methods can be applied to these MILP models to allow realistic cases to be optimized. In this paper we present a Multi-SKU decomposition algorithm based on decomposing the full scale MILP model into submodels covering only a few SKUs. For a small example, this Multi-SKU algorithm could reduce the costs by 1.9% compared to a Single-SKU algorithm. We also apply a Shrinking Horizon (SH) approach to reduce the required computational time. For an example containing 100 SKUs, the Multi-SKU algorithm with submodels containing 5 SKUs could not be solved within 4 hours. Using the SH approach, a solution was obtained in 20 minutes.