Prescriptive Analytics by combining Data Mining and Optimization

What is it about?

We find products that are bought together, using data from multiple stores, to gain a global intelligence of sales patterns using data mining techniques. We then use the sales patterns to formulate mathematical optimization routines. The sales patterns help us to decrease the computational size of the optimization, and make it more accurate at the same time. The optimization helps us to prescribe the correct assortment of products for each individual store, to increase its revenue. Our prescriptive model of product assortment looks at products where price and brand is not a main determinant for sales, but availability of the correct sets of products in a given store is key to their sales. Our model also helps to identify new locations for possible store expansion.

Featured Image

Why is it important?

The market penetration of a given store in a given area depends on many factors, and all stores cannot be treated equally, We develop metrics to identify the growth potential of stores based on its surrounding demographic and other factors. This allows a central planner to accurately estimate growth potential of each store, not based on simplistic revenue projections, but by taking into account the market penetration possibilities of each store. This provides a more realistic reward structure for sales managers of each store. Our model was validated using data from a market leading national retailer with over $500 million revenue in this area.

Perspectives