Most education agencies have implemented new teacher evaluation systems that promise to improve teacher performance. Post-observation performance feedback is a theoretically important driver of this promise as it should ultimately develop teacher-specific weaknesses. This is the first large-scale study to use the written feedback provided to early-career teachers during formal post-observation conferences and quantitatively link critical feedback characteristics (CFCs) to measures of teacher human capital. We find that most conferences do not include CFCs, that feedback is typically unidimensional, and that less effective early-career teachers receive higher shares of CFCs. However, goal-setting is the only CFC associated with subsequent teacher performance. Beginning and less-educated teachers, for whom goal-setting may clarify performance expectations, drive this relationship.