The rise of the evidence-based approach has influenced how criminological research is conducted in Britain. Funding bodies increasingly specify a quantitative research design as the most appropriate way to determine ‘what works’. Indeed, there is a growing tendency to regard the randomised control trial (RCT) as the ‘gold standard’ for such studies. In this article, restorative justice research is used as an example of the pitfalls of relying on a purely quantitative paradigm. By ignoring issues of programme integrity and context, RCTs often produce contradictory or inconclusive results. The article  concludes that the artificial divide often imposed on researchers between qualitative and empirical approaches is unhelpful, and that research into  restorative justice (and other areas) would gain from the integration of an in-depth, qualitative component within the experimental approach.