Using a dataset in which we observe the specific individuals in a household making choices, we study differences in purchase behaviors within and across households in five grocery categories. We develop a choice model that recovers utility parameters from individuals’ choices while allowing for the behavior of individuals from the same household to be correlated via a Bayesian hierarchy. We find that our modeling approach is better at recovering the utility parameters corresponding to each household member’s brand choices especially when we observe few purchase occasions from each of them. The improvement stems from using the household information as a prior for the individual. We find that the within-household heterogeneity in estimated brand intercepts and (to a lesser extent) price sensitivities is about 50% of the across household heterogeneity in these parameters. However, with promotion sensitivities we find within-household heterogeneity to be as large as the across household heterogeneity. We use these estimated utility parameters to compare the expected profitability of promotions targeted at the individual rather than at the household level and find substantial improvements in returns to supermarket promotions. We check the robustness of our results to an alternative model specification and to accounting for possible price endogeneity.