This dissertation covers two topics within the context of online platform design. In chapter 1, I develop a structural model to evaluate the effect of providing user traffic information to heterogeneous bidders in sponsored search auctions. Many internet platforms use sponsored search auctions as their primary source of revenue. In these auctions, advertisers bid for slots with different desirabilities. A standard assumption is that bidders know the click through rate (CTR) for each slot. I relax this assumption in two ways. First, I allow bidders to receive private signals about the CTR of the highest advertising slot, which they use to update their beliefs during an auction. Second, I allow for bidders to start each auction with different priors. I estimate my model using closed-form formulae and a new dataset from a large internet platform where the CTR for the highest slot varies significantly across auctions and over time. My estimates imply considerable variation in bidders' priors, and that this affects platform revenues. Specifically, I predict that the platform's revenues would increase by an average of 7% if the platform was able to credibly and accurately reveal the CTR of the highest slot. I show how this gain in revenues relates to changes in revenue from bidders who, in the absence of knowledge of the CTR, have either optimistic or pessimistic priors about the CTR.In chapter 2, I show how to calculate the theoretically optimal reserve prices in auctions for online advertisements with endogenous platform user behavior. In the case where the advertised content is useful to the platform's users, showing less advertisements due to increased reserve prices could imply less clicks on each advertisement from users because of a smaller choice set. Qualifying more bidders by lowering reserve prices creates a positive externality for all participating bidders. I present the results of a large field experiment in a sponsored search setting. Consistent with the theory, platform revenues increase substantially after the introduction of the optimal reserve prices, while users engage more with the website. In chapter 3, I discuss the economic benefits of a Central Dispatcher for the New York City taxi industry. Drivers make dynamic spatial decisions without taking into account that their decisions impact their fellow drivers and consumer demand, increase traffic and affect matching efficiency. The Central Dispatcher internalizes that driver decisions affect the outcomes of other drivers and have an effect on congestion, demand, and matching efficiency. The Central Dispatcher makes decisions in an environment with search frictions, while taking into account the aforementioned externalities in order to maximize the market's social surplus. I solve for the Central Dispatcher allocation using a value function approximation algorithm based on neural networks. Results indicate that the competitive equilibrium leads to imperfect coordination between the drivers, excess supply and more traffic congestion than the optimum. The Central Dispatcher increases social surplus by 15%, or $798,000 per shift, reduces congestion by 5% on average and increases market thickness in Lower Manhattan and the Boroughs.