Consumers play a central role in solving agency problems. For example, someone who receives poor service in a restaurant or store can ask to "see the manager". Similarly, a person incorrectly denied unemployment benefit or unlawfully arrested has opportunities to appeal the decision to a higher authority. Consumers feedback to firms comes in a variety of forms, such as customer satisfaction surveys, appeals, suits, complaints machanisms, "focus groups", market research surveys etc., and is used for a host of reasons. 1 One reason to solicit (or perhaps a byproduct of soliciting) consumer feedback is to effectively monitor the actions of employees. The purpose of this paper is to understand how well they do so.Consumers are a relatively cheap way of getting information on the performance of employees.Yet agency theory has largely ignored their role: instead, monitoring precision typically appears as an exogenous parameter of the agency problem. (See Prendergast, 1999, for a review.) I argue here that a careful focus on consumers can help understand the ability of firms to resolve agency problems, where their role depends on the congruence in preferences between consumers and principals. 2The idea can be easily explained by a pair of examples. First, consider a restaurant customer who receives poor service. Both the customer and the owner of the restaurant likely wish that the customer has good service, as the restaurant's success depends on it. I show below that in this situation, the agent can be effectively monitored by the consumer, as the consumer has the "right" objectives. In effect, the incentives of the customer are aligned with those of the principal, which both improves incentives and oversight to correct mistakes. By contrast, now consider an applicant for unemployment benefits. Here, there is considerable divergence in the preferences of the Department of Labor and an applicant for unemployment benefit: the applicant always wants to be approved for benefits, while the Department wishes to deny unqualified candidates. I show below that in this latter case not only are customers less useful, but more surprisingly, using consumer feedback can be harmful to efficiency.The model below has the following features. First, a good has to be allocated to a consumer, where the optimal allocation depends on some information available to the agent. Second, she can exert effort to improve the quality of her information. Third, monitoring by the principal is costly, so he would like some information that an error has occurred before intervening. This is where 1 Such feedback is useful to identifying consumer preferences, pointing out defects in product design, correcting inadvertant mistakes, and so on.2 See Aghion and Tirole, 1997, for other work stressing the importance of congruence in preferences, though in their case they emphasize correlation in preferences between agents and principals.