At the University of Teacher Education Vaud, Switzerland, the quality of teaching is assessed by selecting a sample of modules every semester and inviting all enrolled students to evaluate them by completing an online questionnaire. Based on these data, the direction asks that an annual result is computed for the whole institution. How to calculate this overall score, so that it accurately mirrors the students’ voice? Viewing student ratings as data rather than as evaluations, we point out that the chosen statistical model underlying the calculations has considerable influence on the results and thereby on the actions to be taken.