There are inherent differences in the priorities of academics and policy-makers. These pose unique challenges for teams such as the Behavioural Insights Team (BIT), which has positioned itself as an organisation conducting academically rigorous behavioural science research in policy settings. Here we outline the threats to research transparency and reproducibility that stem from working with policy-makers and other non-academic stakeholders. These threats affect how we perform, communicate, verify and evaluate research. Solutions that increase research transparency include pre-registering study protocols, making data open and publishing summaries of results. We suggest an incentive structure (a simple ‘nudge’) that rewards BIT's non-academic partners for engaging in these practices.