We propose to extend the data broker system concept and perform empirical experiments and lab experiments using human subjects. Our empirical and lab experiments will test users’ strategies and behavior with respect to their bidding as well as their willingness to share their personal data. For theoretic extension toward more practical online advertising systems, we propose to study online advertising market mechanism design in a repeated setting where users, mediators, and advertisers could be no-regret learners, and ask: how should a mechanism do to (approximately) maximize the gain over time?