We study a model in which an online publisher makes a profit by auctioning an advertising slot that appears whenever a consumer visits its website. Consumers are heterogenous in terms of the products they like, and the publisher has gathered data which are statistically correlated with consumers' tastes. The optimal disclosure policy for the publisher solves the trade-off between rent-extraction and dispersion of bidders' willingness to pay. Disclosure of consumers' data improves the average quality of a match, but also leads to higher prices.