Successful decision-making for environmental management requires evidence of the performance and efficacy of proposed conservation interventions. Projecting the future impacts of prospective conservation policies and programs is challenging due to a range of complex ecological, economic, social and ethical factors, and in particular the need to extrapolate models to novel contexts. Yet many extrapolation techniques currently employed are limited by unfounded assumptions of causality and a reliance on potentially biased inferences drawn from limited data. We show how these restrictions can be overcome by established and emerging techniques from causal inference, scenario analysis, systematic review, expert elicitation, and global sensitivity analysis. These technical advances provide avenues to untangle cause from correlation, evaluate and transfer models between contexts, characterize uncertainty, and address imperfect data. With more rigorous projections of prospective performance of interventions, scientists can deliver policy and program advice that is more scientifically credible.