In the restaurant industry, delivery platforms collect customer orders via mobile apps or the web, transmit them to restaurants, and deliver the orders to customers. Platforms claim they provide value to restaurants by expanding their markets, but critics claim they destroy restaurant profitability by taking a percentage of revenues from delivery orders and generating congestion that negatively impacts dine-in customers. We consider these tensions using a stylized model of a restaurant as a congested service system, with a goal of understanding how best to structure relationships between restaurants and platforms to maximize joint profits. First, we analyze the efficacy of the most commonly used contracts, “simple revenue sharing,” in which the platform takes a percentage cut of each delivery order. We show that such contracts fail to coordinate the system because the platform does not internalize the effect of its price on dine-in revenues, setting its price too low and attracting many delivery customers who generate congestion for dine-in customers; the restaurant is forced to respond to this by setting dine-in prices that are too low, which in turn negatively impacts restaurant revenues compared to a centralized system in which both prices are set to maximize joint revenue. We then investigate several proposed mechanisms to improve the relationship between restaurants and platforms and coordinate the system. We show that commission caps, which have been legally mandated in some US states and localities, are ineffective at coordinating the system because they do not solve the structural problem with simple revenue sharing contracts. Allowing the restaurant to set price floors for the delivery platform is similarly incapable of coordination, generating double marginalization that leads to prices that are too high relative to the central optimum. Finally, we propose an alternative, practical coordinating contract that is a variation of simple revenue sharing: for each delivery order, the platform pays the restaurant a fixed, flat fee and a percentage revenue share. For appropriately chosen contract parameters, we show that this contract coordinates the system, protects restaurant margins by ensuring a lower bound on its revenue per delivery order, and allocates revenue with a high degree of flexibility between the restaurant and the platform.