In many economic situations several principals contract with the same agents sequentially. Asymmetric learning about agents' abilities provides the first principal with an informational advantage and has profound implications for the design of incentive contracts. We show that the principal always strategically distorts information revelation to future principals about the ability of her agents. The second main result is that she can limit her search for optimal incentive schemes to the class of relative performance contracts that cannot be replicated by contracts based on individual performance only. This provides a new rationale for the optimality of such compensation schemes.