We propose a model where a privately informed agent submits an application to obtain a benefit, such as a patent approval, a journal publication or a research grant. The evaluator observes a noisy signal about the applicant’s merit and decides whether to approve or reject the application. The application serves as a costly ordeal, encouraging self-selection of applicants more likely to succeed. What is the optimal ordeal the evaluator should impose on the applicant? When is it optimal to make the application costless? We characterize the optimal ordeal and identify monotonicity of the hazard ratios (ratio of hazard rates) of the applicant’s and evaluator’s signals as the driving condition for the results.