This paper estimates the impact of college teaching on students' academic achievement and labor market outcomes using administrative data from Bocconi University (Italy) matched with Italian tax records. The estimation exploits the random allocation of students to teachers in a fixed sequence of compulsory courses. We find that good teaching matters more for the labor market than for academic performance. Moreover, the professors who are best at improving the academic achievement of their best students are also the ones who boost their earnings the most. On the contrary, for low ability students the academic and labor market returns of teachers are largely uncorrelated. We also find that professors who are good at teaching high ability students are often not the best teachers for the least able ones. These findings can be rationalized in a model where teaching is a multi-dimensional activity with each dimension having differential returns on the students' academic outcomes and labor market success.