“Building Optimized and Hyperlocal Product Assortments: A Nonparametric Choice Approach used by Celect” by Vivek Farias (MIT), Srikanth Jagabathula (NYU), and Devavrat Shaw(MIT).
Bricks-and-mortar retailers with multiple stores must address the local tastes and preferences of the customers in those different locations. For each store location, and given a dollar budget, the firm must decide how many of each product, style and color to carry—the assortment problem. The authors bring a new methodological approach to this assortment problem, using individuals’ historical shopping patterns to predict demand for required stock keeping unit levels for specific stores. They report an improvement of 6.65% in revenue when the approach was implemented, in line with model forecasts.