This paper reflects on a process evaluation of a home visitation programme in South Africa. The programme, implemented in two low-income communities, focused on the reduction of risks to unintentional childhood injuries. The evaluation comprised a combination of qualitative and quantitative methods, including observations in conjunction with an evaluator's journal, diaries kept by the home visitors, interviews and focus group discussions. Short questionnaires were administered to programme staff and home visitors. Caregivers were visited to attain their assessment of visitors and the programme. These methods resulted in a detailed description of implementation processes, but more importantly gave insight into the experiences and perceptions of the social actors, i.e. programme staff, visitors and caregivers. It also offered possible explanations for the difference in the intervention effect between the two sites. Two major challenges to the evaluation were: (i) the power-imbalance between the evaluator and community participants (visitors and caregivers) and (ii) the language- and cultural barriers between evaluator and community participants. The evaluation demonstrated that process information can contribute towards explaining outcome results, but also that active participation from all social actors is a necessary condition if process evaluations are to result in programme improvement. (Contains 3 tables, 2 graphs and 4 boxes.)