Abstract:

 The assortment optimization problem is a classical problem in revenue management. In this problem, the retailer has to determine the subset of products to offer from a much larger set, so as to maximize the expected revenue subject to operational constraints. The core of this problem is how to characterize customers' choice behavior among differentiated products, develop optimization models, and design efficient solution algorithms. In this paper, we review existing studies on assortment optimization problems under logit-based discrete choice models. We first introduce assortment optimization problem based on the multinomial logit model. Next, we cover two advanced variants: (1) The first variant is based on the two-leve or multi-level nested logit models, which are able to take into consideration the substitution effects among differentiated products; and (2) The second variant is based on the mixtures of multinomial logits model, which can capture the heterogeneity among customers. Then, we cover the data-driven assortment optimization problem under rank-based non-parametric model. Finally, we outline possible directions for future research.

Key words: assortment optimization, revenue management, multinomial logit, nested logit