This paper models two key roles of subjective performance evaluations: their incentive role and their feedback role. The latter is important in real world firms but largely overlooked by economists. The paper shows that in contrast to received wisdom, interaction between these two roles makes subjective pay feasible even in a finite horizon relationship. The optimal contract depends critically upon the existence and quality of a verifiable performance measure. Furthermore, when commitment to a forced distribution of evaluations is possible, it is valuable only if the firm employs many workers or if the verifiable measure is poor.