With the abundance of online big data and computing resources, as well as the rapid development and diversification of business models, programmatic advertising (PA) has emerged and become the mainstream and one of the most promising advertising channels in recent years. As an effective way to precision marketing, PA relies heavily on the information elicited from analyzing Web users and pages, which can help advertisers precisely identify their best-matched audiences and evaluate the ad impressions in a real-time fashion. In PA markets, publishers serve as the supplier of ad impressions, and have control on whether, what and how to reveal such key information to advertisers. These decisions play a central role in the information structure of data-driven PA markets, and attract intensive research interests. In this paper, we strive to investigate publishers’ rational preference over the symmetric and asymmetric information structures, with the aim of maximizing their revenues. Our model views publishers’ revenue as a function of the information structure among advertisers, and we hereby proved its convexity under an incentive compatible mechanism. This conclusion indicates that publishers prefer an asymmetric information structure rather than a symmetric one. Our research findings can help improve publishers’ revenue, and also can explain the underlying rationale for the hybrid PA markets mixed by public real-time bidding and private marketplaces.