Districts nationwide have increased the frequency of teacher evaluations. Yet, we know little about the role of evaluator feedback for teacher improvement. Using unique classroom observation-level data, we use evaluator ratings and teacher self-assessments of teacher performance to rigorously examine (positive and negative) feedback valence from the teacher's perspective and its consequences. While teachers and evaluators similarly rate performance, there's significant variability in teacher-evaluator ratings. Teacher performance improves across multiple within-year classroom observations, though evaluator ratings likely overstate improvements among the lowest-performing teachers. While negative feedback from evaluators likely improves within-year teacher performance and may improve their productivity regarding student achievement, statistically insignificant yet practically meaningful evidence suggests it may also push teachers toward schools with more positive feedback.