The early stages of many careers serve as a trial period where employees seek a high-value reward (e.g., a promotion) while working hard to demonstrate their productivity. A key design component of such jobs is the performance feedback offered to the employee, as it allows them to adjust their behaviors and learn what their future rewards might look like. While some experts argue that a policy of full transparency is best---that is, keeping employees fully apprised of their performance---such practice is far form uniform as employers may see a strategic gain from hiding information or postponing its release.

We study an agency model in which the principal has outcome data under different incentive schemes and aims to design an optimal contract under minimal assumptions about the way the agent responds to incentives. Events unfold as follows: (1) the principal offers a contract---a mapping from output to nonnegative payments; (2) the agent chooses costly action---a probability distribution over output; and (3) output and payoffs are realized. The principal has outcome data under K different contracts which, sidestepping estimation error, enables her to recover the action corresponding to each of these contracts. We assume that the agent best-responds to the offered contract and has quasi-linear preferences over money and actions, but we make no further assumptions about the production environment. The principal does not have prior beliefs about any of the unknown aspects of the environment. Instead, she seeks a contract that maximizes worst-case profit.

We derive an optimal dynamic contest for environments where effort can be monitored only through a coarse, binary performance measure and the principal chooses prize-allocation and termination rules together with a real-time feedback policy for the contestants. The optimal contest takes a stark cyclical form: contestants are kept fully apprised of their own successes, and at the end of each fixed-length cycle, if at least one agent has succeeded, the contest ends and the prize is shared equally among all successful agents irrespective of when they succeeded; otherwise, the designer informs all contestants that nobody has yet succeeded and the contest resets. Applications include promotions, innovation contests, and proof-of-work protocols.