The Value of Personalized Recommendations: Evidence from Netflix

Personalized recommendation systems shape much of user choice online, yet their targeted nature makes separating out the value of recommendation and the underlying goods challenging. We build a discrete-choice model that embeds recommendation-induced utility, low-rank heterogeneity, and flexible state dependence and apply the model to viewership data at Netflix. We identify recommendation-induced engagement from observational variation in algorithmic exposure, treating exposure as conditionally exogenous given observed user, good, and state characteristics. Separately, we estimate model-free diversion ratios from a randomized experiment that perturbs recommendations and use these to assess the validity of the structural model. We use the model to evaluate counterfactuals that quantify the incremental engagement generated by personalized recommendations. First, we show that replacing the current recommender system with a matrix factorization or popularity-based algorithm would lead to a 4% and 12% reduction in engagement, respectively, and decreased consumption diversity. Second, most of the consumption increase from recommendations comes from effective targeting, not mechanical exposure, with the largest gains for mid-popularity goods (as opposed to broadly appealing or very niche goods).