This paper discusses the spectrum of synthesis methods available to generate, explore and text theory, their value to the field of international development and innovations required to make better use of the primary research available. It argues for clearer distinctions between syntheses produced as public goods, and those tailored to specific circumstances; and strengthening knowledge systems through greater use of maps to navigate existing and missing evidence, harmonised outcomes and measures, and advances in automation technologies. Improved methods and guidance are required for synthesising formative research and investigating contextual factors. Engaging stakeholders and working across academic disciplines support the production of policy-relevant syntheses and inspire methods development. This paper was prepared during the early stages of setting up DFID's Centre for Excellence in Development Impact and Learning (CEDIL), which was commissioned to innovate in the field of impact evaluation; design, commission and implement impact evaluations; and promote the uptake and use of evidence from impact evaluations. We drew on discussions within the CEDIL Intellectual Leadership Team and are particularly grateful to Nancy Cartwright, Rick Davies, Andrew Pullin and Howard White for their comments on earlier versions of this paper.