Inferring Consideration Sets from Sales Transaction Data

Abstract

Understanding consumer preferences is critical when optimizing prices and assortments in retail operations, and when matching supply and demand in online platforms. In pursuing such an objective, a key input is the set of products that are both available and considered, from which a customer makes a choice. In practice, product availability is often hard to predict because it is subject to market conditions or other operational factors and considered products are unobserved.} In this paper we propose a methodology to identify consideration sets, defined as those that are available and considered, from sales transaction records in a data driven way. We assume that customers are boundedly rational and make purchases in a two-stage process. First, they sample their consideration set and then purchase the most preferred item therein. Our contribution to the literature is two-fold. Theoretically, we address the problem of identifiability of consider-then-choose models from data. Since calibrating this class of choice models is a hard problem, we propose a framework to effectively estimate them and infer consideration sets. The methodology to model the consideration set formation is founded on machine learning techniques that can account for nonlinear-in-parameter utilities in a tractable way. Then we apply the proposed framework on synthetic data and two real datasets: one from a big retail chain and the other one from a car-sharing platform. We observe that accounting for consideration sets can significantly boost the predictive performance in comparison with classical choice-based demand benchmarks, particularly in cases when the assortment of available products is not clearly defined. Our findings suggest that consider-then-choose models tend to be rather robust to the degree of ambiguity in the consideration set definition, and their relative performance with respect to state-of-the-art choice-based demand models improves with this ambiguity.