Physical inactivity is a significant driver of health and social inequalities, particularly affecting socioeconomically disadvantaged communities. This poses a major challenge to policymakers worldwide. Despite the large volume of original research and reviews that focus on the design and evaluation of interventions to increase physical activity, there remains little consensus on which interventions are likely to work. This paper discusses physical activity interventions through the lens of behavioural science. We consider the conclusions drawn by previous reviews of this literature and link them to potential behavioural mechanisms that might explain them. We categorise interventions into three broad types: physical environment, information provision and social context, and discuss specific components within each that are known to influence behaviour. The paper is not a systematic nor an exhaustive review. The recommendations are not for implementation without testing. Rather, the paper contributes an analysis of how existing evidence can be used to design research and interventions in future to test not just the main outcome, but the behavioural mechanisms that may determine success.