This paper is motivated by recent initiatives in which e-commerce platforms such as Shopify are inviting merchants to pool their audiences, creating a new kind of market where customer data are traded for targeted advertising. Pooling audiences can improve targeting but exposes merchants to customer poaching. I develop a model of targeted advertising in which each customer has merchant-specific click-through rates and study how a platform can incentivize merchants to pool audiences despite poaching risk.  In a sufficiently large (limit) market, the optimal mechanism offers merchants three options: (i) receive, at no charge, a campaign built on the pooled audience that delivers the same expected clicks as advertising to their own audiences, (ii) expand it by paying a posted buying price per additional click, or (iii) reduce it and receive a lower posted selling price per click forgone. Both prices are determined by endogenous virtual prices, which equal both the (weighted) marginal revenue and marginal cost of selling and procuring a click, respectively. Virtual prices also determine which customers are targeted by which merchants, through a comparative-advantage rule that compares relative rather than absolute click-through rates.

A novel feature of the optimal mechanism is the exchange option (i), in which a merchant neither makes nor receives payments but trades his own customers for different ones. The paper's main technical contribution is a convergence result showing that, as the number of merchants grows, an optimal finite-market mechanism converges to the optimal limit-market mechanism.

This paper is motivated by the observation that experts (e.g., consulting firms) charge value-based fees for some services and flat fees for others. I study how an expert should price her services when clients privately know which project they are interested in and how much they value expertise on it. In my model, expertise provided on one project can be valuable for a client interested in a related project. This creates a new incentive problem: a client may buy cheaper advice meant for another project if it provides a sufficiently accurate insight into his true project of interest. I find that when spillovers are sufficiently weak, value-based fees are optimal, and the expert provides full information and extracts the full surplus by charging each client his willingness to pay for it. When information spills over perfectly across projects, a common flat fee for full information is optimal, with higher-value clients enjoying more surplus. Between these extremes, the expert distorts the information she provides to low types and extracts their full surplus, but gives full information to middle and high types. Because higher-value clients lose more from less relevant advice than they save in fees, the expert can charge them higher fees relative to their willingness to pay, so surplus can peak among the middle types and fall to zero at the top.

The results rationalize pricing, product design, and patterns of client surplus that cannot arise in standard screening models. The paper’s main technical contribution is a Markov model of information spillovers that permits a full characterization of the optimal mechanism.

(extended abstract and/or preliminary draft available upon request)