Random Utility Models with Cardinality Context Effects for Online Subscription Service Platforms

Abstract

A more general family of random utility models is developed to model a cognitive heuristic, known as consideration sets. These new models, denoted as Multinomial Logit Cardinality Effect models (MNL-CE), define perceived representative utility of items by assigning a penalty as a function of assortment cardinality to the representative utility of each item beyond a threshold value (except for the no-choice option). This definition of perceived representative utility of an item is context-dependent and thus a function of assortment attributes (cardinality), in addition to item and user attributes. The user’s net benefit is therefore a trade-off between the benefits and the costs of considering a certain number of items. A developed algorithm efficiently solves the subscription platform assortment optimization problem with equal profit when user selection is modeled via variants of the MNL-CE. The sensitivity of model parameters on the optimal assortment cardinality and no-choice probability is analyzed with the MovieLens dataset.