Implemented bayesian machine learning models to estimate advertiser valuations on ad auctions using publicly available data from Twitter. Used the results to inform an auction-theoretic model of online ad auctions to answer various policy questions.