Store-level assortment optimization

A company specializing in the sales of luxury accessories wanted to reconsider the clustering of its points of sales according to different characteristics of these points of sales as well as their customers’.

The goal was then to be able to explain in each cluster the parameters that affect the sales of a specific range of products, and to extract recommendations to increase sales in the future. The purpose was to provide a reliable projection of future sales according to the parameters chosen for each cluster.


  • Analyze similarities between points of sales through different axes to create relevant clusterings
  • Quantify the impact of each parameter so that consistent revenue forecasts can be made using only the most important variables.
  • Build models (regression models) to change attribute values and observe changes in the underlying sales projection for future years.


  • Provide a model and tool to forecast sales according to a product assortment (volume and variety of SKUs in store inventory)
  • Define clusters of stores for each product family