The choice of research method relevant to the evaluation of delivery of a health intervention is not always straightforward. We use the evaluation of HIV and tuberculosis community treatment supporters in promoting adherence to treatment in Africa as a case study to illustrate the pros and cons of operational research and randomised controlled trials. The choice of this intervention for the case study reflects the importance of maximising the benefits of unprecedented efforts to scale-up treatments of these two epidemics. International policy supporting the role of community treatment supporters in tuberculosis is largely based on the findings of operational research studies. This reflects the advantages that operational research is less costly than randomised controlled trials, provides more rapid answers to policy questions, enables standard evaluation of the intervention in 'real life' conditions in several diverse settings and has in-built potential to influence policy and practice, because the research is conducted within health programmes. Recent evidence on the role of community treatment supporters in HIV is largely based on randomised trials. This reflects the advantages that randomised trials compared to operational research are more rigorous and generate a more convincing result. Operational research and randomised trials may be viewed as providing complementary findings to inform new policies and practice aimed at improving programme performance and patient outcomes. However, in practice, insufficient funds are likely to be made available for randomised trials to answer all the current research questions on delivery of programme interventions. In deciding on the type of research to evaluate a particular health intervention, dialogue is necessary with policy-makers to weigh up explicitly the trade-offs between research rigour and other factors such as cost, speed of implementation of research and speed of policy uptake and of change in programme practice.