Using a new longitudinal study of UK undergraduate students, we estimate the production function of graduate skills at university, which we proxy with marks. We use a flexible functional form and a large set of inputs. The inputs considered include both students’ investment (time spent attending lectures and classes, study time spent using active learning techniques, study time spent using passive learning techniques) and students’ characteristics (past educational attainment, cognitive skills and non-cognitive skills). We deal with the endogeneity of the student’s investments by allowing unobserved heterogeneity in outcome and investment to be correlated and by exploiting random class assignment. We find that attendance to lectures and classes and study time are important inputs in the production function of graduate skills. However, time spent using active learning techniques is much more productive than time spent on passive learning techniques. Interestingly, we find that cognitive skills are important determinant of students’ investment, but have no direct impact in the production function. The reverse is true for cognitive skills.