We propose a computationally simple semiparametric discrete choice estimator to model rich consumer heterogeneity. We assume groups of observably similar consumers have similar preferences, but allow preferences to vary freely across these groups. Model flexibility is easily adjusted by setting a single tuning parameter; we suggest a cross‐validation method to do so. We analyze the model's properties in the context of hospital mergers, both analytically and via a Monte Carlo analysis. The model performs well for policy relevant substitution and welfare measures, even if misspecified, when the tuning parameter is set within the neighborhood of the value chosen by cross validation. (JEL C14, D12, I11, L41)