Rigorous impact evaluations, particularly randomised trials, can provide governments with valuable insights into whether policies and programmes achieve their intended outcomes. Employed effectively, they offer an evidence base that goes beyond assumptions and precedent, supporting better resource allocation and more effective services for citizens. However, despite their potential, such evaluations remain underused in many contexts, given a number of technical and political barriers. This report explores how governments can overcome these barriers and deliver high-quality evaluations to contribute to policy development. It briefly discusses the potential for artificial intelligence to contribute to impact evaluation. It sets out the main evaluation methods available, ranging from randomised trials to quasi-experimental approaches, and highlights the conditions under which each can be applied. It addresses ethical issues and highlights how such concerns can be addressed through careful design and stakeholder engagement. There are also options for ensuring that evaluations can remain cost-effective, including through greater use of administrative data, alignment with policy priorities, and partnerships with wider networks. In particular, it underlines the value of international co-operation and peer learning to build capacity, share methods, and upscale effective programmes.
