SMU SOE Seminar Series (January 26, 2026): Managing by Feedback

TOPIC:

ABSTRACT

This paper studies how managers influence worker productivity through feedback. Using data from GitHub and LinkedIn, I analyze over 200 million pieces of feedback during code reviews across 1.7 million software teams. I apply large language models to classify feedback by tone (toxicity, positivity) and informational content (constructiveness). I exploit random reviewer assignments to estimate the causal effects of feedback on developer productivity and retention. Toxic feedback reduces future code quantity and quality and lowers developer retention within the firm, whereas non-toxic criticism has no such detrimental effects. Positive feedback increases productivity and retention and generates spillovers to coworkers. Constructive feedback does not affect future code quantity, though it lowers quality because revisions to reviewed code crowd out time spent on new code development. Finally, I show that feedback explains a sizable share of variation in manager quality, measured by value added to worker productivity. Overall, this paper shows that feedback tone and information shape worker productivity and retention, offering new insights into effective management.