This paper studies the effect of providing feedback to college students on their position in the grade distribution by using a natural field experiment. This information was updated every six months during a three-year period. We find that greater grades transparency decreases educational performance, as measured by the number of exams passed and GPA. However self-reported satisfaction, as measured by surveys conducted after feedback is provided but before students take their exams, increases. We provide a theoretical framework to understand these results, focusing on the role of prior beliefs, and using out-of-trial surveys to test the model. In the absence of treatment, a majority of students underestimate their position in the grade distribution, suggesting that the updated information is “good news” for many students. Moreover, the negative effect on performance is driven by those students who underestimate their position in the absence of feedback. Students who overestimate initially their position, if anything, respond positively. The performance effects are short lived - by the time students graduate, they have similar accumulated GPA and graduation rates.