BACKGROUND: Obesity is an important public health issue. Finding ways to increase physical activity and improve nutrition, particularly in children, is a clear priority. Our Cochrane review of the World Health Organization's Health Promoting Schools (HPS) framework found this approach improved students' physical activity and fitness, and increased fruit and vegetable intake. However, there was considerable heterogeneity in reported impacts. This paper synthesises process evaluation data from these studies to identify factors that might explain this variability. METHODS: We searched 20 health, education and social-science databases, and trials registries and relevant websites in 2011 and 2013. No language or date restrictions were applied. We included cluster randomised controlled trials. Participants were school students aged 4-18 years. Studies were included if they: took an HPS approach (targeting curriculum, environment and family/community); focused on physical activity and/or nutrition; and presented process evaluation data. A framework approach was used to facilitate thematic analysis and synthesis of process data. RESULTS: Twenty-six studies met the inclusion criteria. Most were conducted in America or Europe, with children aged 12 years or younger. Although interventions were acceptable to students and teachers, fidelity varied considerably across trials. Involving families, while an intrinsic element of the HPS approach, was viewed as highly challenging. Several themes emerged regarding which elements of interventions were critical for success: tailoring programmes to individual schools' needs; aligning interventions with schools' core aims; working with teachers to develop programmes; and providing on-going training and support. An emphasis on academic subjects and lack of institutional support were barriers to implementation. CONCLUSIONS: Stronger alliances between health and education appear essential to intervention success. Researchers must work with schools to develop and implement interventions, and to evaluate their impact on both health and educational outcomes as this may be a key determinant of scalability. If family engagement is attempted, better ways to achieve this must be developed and evaluated. Further evaluations of interventions to promote physical activity and nutrition during adolescence are needed. Finally, process evaluations must move beyond simple measures of acceptability/fidelity to include detailed contextual information to illuminate exactly what works, for whom, in what contexts and why.