Contests, in which contestants compete for a prize offered by a contest holder, have become a popular way to source innovation. Despite great interest from the academic community, many important managerial aspects of contests have received very little formal inquiry. The most important of these is feedback from the contest holder to the contestants while the contest unfolds. This paper sets out to establish a comprehensive understanding of how to give feedback in a contest by answering the questions of when to give feedback and when not to give feedback and which type of feedback to give, public (which all solvers can observe) or private (which only the concerned party can observe). We find that feedback will not affect the behavior of competing problem solvers unless the contest holder credibly precommits to a truthful feedback policy. We then set up a framework that reduces the feedback decision to a pair of conceptual questions. First: Is the contest's ultimate objective to increase average quality or to find the best solution? Second: How uncertain are outcomes for the solvers? We show that no feedback or public feedback generally dominate private feedback. However, if the host is interested exclusively in the best performance and if the contest displays large uncertainties, private feedback is optimal.