Social scientists have often had difficulty evaluating the impact of probation services, partly because expectations and political circumstances change and partly because appropriate methodologies have been slow to develop. This chapter outlines the history of evaluative research on probation. It describes the limitations of early probation research which led to erroneous conclusions that ‘nothing works’, and goes on to show how more recent research has been based on a fuller understanding of practitioner inputs through research on programmes, skills and implementation. This is starting to lead to a better understanding of which practices are effective (‘What Works’). The chapter advocates a mixed qualitative and quantitative methodology for evaluative research which combines understanding, measurement and comparison. Finally, it points to some risks to evidence-based policy which arise from current populism and post-truth politics.