Dynamic assortment in the presence of brand heterogeneity

In this paper, we study the dynamic assortment planning problem in the presence of heterogeneous brands. Over a limited selling season, the retailer sells heterogeneous products from one store brand and one national brand. We use the nested multinomial logit (NMNL) framework to model consumer choice process, in which consumers choose first which brand to buy and then a product within that brand. We formulate this problem using the finite-horizon dynamic programming approach. Using available sales transaction data, we estimate the consumer choice behavior and empirically demonstrate existence of brand heterogeneity. Further, our results suggest that ignoring brand heterogeneity will make the retailer’s expected revenues significantly overestimated, and the potential revenue overestimation depends on initial inventories and prices of products of two brands. We finally show that the retailer will benefit from dynamic assortment optimization with the estimated consumer choice model.