Purpose. This study assessed the short‐term impact of offending behaviour programmes in relation to certain key features of programme delivery identified by an accreditation system. The aim was to further inform the debate on ‘What Works’ in practice by establishing whether well‐delivered programmes are more effective.Method. The sample consisted of 5,255 offenders serving custodial sentences in prisons across England and Wales who completed one of two accredited cognitive skills programmes during the financial year 2001/2002. The relationship between a battery of assessment measures, tutor experience, drop‐out rates, audit observations and the quality of programme delivery was explored.Results. A positive short‐term impact was observed on the majority of assessment measures across both programmes. Patterns of change were broadly similar across gender, age, and ethnic groupings. This short‐term impact of the programmes was significantly greater for a group of high‐need prisoners and at those sites where tutors were delivering more frequently. Tutor delivery rates were also found to be related to drop‐out rates and the quality of programme delivery. Attempts to establish a relationship between ratings of tutor performance from video‐monitoring and the short‐term impact of the course were unsuccessful. However, at the programme site level, positive correlations were observed between the quality of delivery and other measures of the site's performance.Conclusion. This study shows that programmes do have a short‐term impact and that this is greater for higher‐need prisoners and at sites where tutors were delivering more frequently. Furthermore, the accreditation system has highlighted key aspects of programme delivery which do appear to influence the short‐term effe