Research — Kohei Onzo

Sequential search models are popular in marketing for studying consumer search behavior. Current search models use parametric assumptions regarding different aspects of the models, such as random shocks, search costs, and consumer preferences. These assumptions can be restrictive. The authors develop a novel Bayesian nonparametric framework for sequential search to flexibly model unknown distributions. They also develop Markov chain Monte Carlo methods for inferring the model unknowns. The authors conduct simulation studies to demonstrate that currently popular parametric search models can yield incorrect estimates and inferences when the data-generating process deviates from their assumptions. In contrast, the proposed model accurately recovers these quantities for various data-generating processes. The methodology is then applied to online search and purchase data from a Japanese retailer. The results show that several heterogeneity distributions have complex patterns, such as multimodality and skewness, that are not captured by a parametric benchmark model. The authors also estimate the monetary value of search costs, the price elasticities, and a counterfactual profit gain under a personalized couponing strategy and find substantial differences in the results from the two models.

Online stores typically offer product filters to help consumers narrow their search. In some contexts, these filters overlap so that the same product appears across multiple filters. Consumers can also vary in how they search: some apply several filters while others purchase directly from a listing page without search. We propose a model that accommodates overlapping filters and heterogeneous search behavior as a dynamic program. The resulting model generalizes the Weitzman framework and other approaches that yield index-based policies. As existing estimation methods are not suitable for handling our large state space, we develop a novel Bayesian MCMC method based on inverse reinforcement learning that uses neural network-based value function approximation for estimation. We apply our model to clickstream data from an online retailer of Japanese incense. Our results show that search costs increase substantially with row position and that the variability of the post-search shocks is greater than that of the pre-search shocks. In counterfactual simulations, consolidating price filters shifts the composition of filter selections toward non-price filters. We also find that prohibiting purchases directly from listing pages reduces the predicted inside-good purchase probability by roughly seven percentage points.

Understanding and predicting consumer preferences in dynamic markets is crucial for marketing decision-making. Existing methods for modeling dynamic consumer heterogeneity are either inflexible in representing individual-level preference trajectories or as in the case of recently proposed Gaussian process-based approaches, induce unimodal cross-sectional heterogeneity distributions. This prevents them from accurately capturing the growth, decline, or divergence of consumer segments over time. We introduce a novel Bayesian nonparametric machine learning framework that leverages a dependent Dirichlet process for modeling heterogeneity to overcome these methodological constraints. Our approach combines the strengths of Dirichlet processes and Gaussian processes to flexibly model a collection of temporally correlated population distributions. This structure simultaneously captures smooth individual-level preference trajectories and evolving multimodal population distributions. Extensive simulations show that our model significantly outperforms a state-of-the-art GP-based benchmark in recovering parameters and predicting choices when the evolving population distributions are multimodal and performs comparably when they are normal. An application to IRI scanner data for multiple categories spanning the Great Recession shows that our model provides more plausible price elasticity estimates, improved predictions, and granular insights into preference shifts during the recession. We finally show that coupon targeting strategies based on our model can generate higher profits.