<p>There is a clear imperative to prevent youth crime and violence owing to the significant social and economic costs for individuals involved and wider society. A considerable evidence base for ‘what works’ in this space exists but gaps remain. To address these, it is necessary to evaluate intervention effectiveness. The aim of this article is to explore challenges and potential solutions regarding outcome measurement in such evaluations. To do this we draw on the literature and our experience as evaluators. We argue that outcome selection should be informed by a robust theory of change, and that normally this will mean focusing on non-crime/violence outcomes. We identify challenges with self-report measures of youth crime and violence, and consequent effects on participants’ engagement in the intervention and the quantity and quality of data they provide. We also discuss ways of using administrative data on crime and violence in evaluations and the practical, scientific and ethical issues these raise. While the challenges identified can partially be addressed, we conclude that currently evaluations of preventive interventions should mostly focus on intermediate outcomes with a known causal link to crime and violence. The longer-term impact on crime and violence can be tracked using administrative data accessed via an archive. We appreciate policymaker and funder pressure to quantify short-term intervention impact on crime and violence but contend that it is important not to force the issue. Meanwhile, it is essential to test the feasibility and acceptability of alternative approaches to measuring crime and violence in evaluations and to share the learning.</p>