I study the impact of consumer reviews on the incentives for firms to participate in the market. Firms produce goods of heterogeneous, unknown quality that is gradually revealed via consumer reviews, and face both entry and exit decisions. A platform combines past reviews to construct firm-specific ratings that help guide consumer search. When the platform integrates all reviews into ratings - full transparency - consumers form queues at the highest-rated firms. This demand cliff induces an S-shaped continuation value for firms as a function of ratings, generating both low entry rates as well as unwanted selection effects - high-quality firms exit early. Whereas firms prefer more feedback when starting out and less feedback when established, equilibrium induces precisely the reverse profile. I then study the design of ratings systems. The platform must balance the need to provide consumers with accurate information against the need to encourage high-quality firms to enter and remain active. The key insight is that optimal rating systems involve upper censorship, i.e. the suppression of reviews from highly-rated firms' ratings, as a means of incentive provision. This makes the task of “climbing the ratings hill” less daunting, stimulating participation. An exploratory calibration using data provided by Yelp! estimates a consumer welfare gain of roughly 7% from adopting the optimal policy.