Field experiments which test the application of behavioural insights to policy design have become popular to inform policy decisions. This study is the first to empirically examine who and what drives these experiments with public partners. Through a mixed-methods approach, based on a novel dataset of insights from academic researchers, behavioural insight team members, and public servants, I derive three main results: Firstly, public servants have a considerable influence on study setup and sample design. Secondly, behavioural insight team members report concerns regarding scientific rigor and limitations imposed by risk-aversion of their public partners significantly more often than academic researchers. Thirdly, transparency and quality control in collaborative research are low with respect to pre-analysis plans, the publication of results, and medium or long term effects. To remedy the current weaknesses, the study sketches out several promising ways forward, such as setting up a matchmaking platform for researchers and public bodies to facilitate cooperation, and using time-embargoed pre-analysis plans.