Using the example of a governance system that allocates public funding for research on the basis of rankings of research quality and impact (as has been developed in the UK over the past three decades), this paper explores three conditions needed for such rankings to be effective as a basis for genuine performance improvement over time. First, the underlying metrics must be capable of meaningfully distinguishing the performance of the institutions being ranked. Second, the basis of assessment must be stable enough for changes in performance over time to be identified. Third, the ranking system should avoid perverse consequences arising from strategic responses by the institutions being assessed. By means of a hypothetical example of a serie...