Identifying and implementing policy measures to reduce on- farm biodiversity loss requires a better understanding of complex system interactions. Studies on policy challenges to enhance on-farm biodiversity can be divided into those focusing on biodiversity-enhancing practices on farms versus those focusing on the success of policies in terms of farmers adoption rates. In this study, we combine both parts within an ad-hoc decision analytic method to predict which policy-scenario is most likely to improve biodiversity on smallholder farms across the European Union. We apply a Bayesian decision network (BDN) to assess on-farm biodiversity outcomes under five policy scenarios including a) regulatory, b) market-based, c) incentive-based, d) voluntary information-driven, and e) policy mix. We collect and combine information from existing literature along with expert elicitation from a workshop held with decision makers and other stakeholders. We found the highest probability of positive biodiversity outcomes (48%) under a regulatory policy but the result was strongly dependent on the expected rate of adoption. We showcase a novel approach combining knowledge from different knowledge strains and sources that accounts for layers of uncertainty while informing biodiversity policy decisions.