This article compares and contrasts two versions of the Education Endowment Foundation's (EEF) Teaching and Learning Toolkit ('Toolkit'), a web-based summary of international evidence on teaching 3-18 year-olds. The Toolkit has localised versions in six different languages in Australia, Cameroon, Chile, Jordan and Spain. The initial Toolkit, created in 2011 with funding from the Sutton Trust and updated since then with funding from EEF, drew upon over 250 meta-analyses across 30 areas of education research. An updated version, drawing on a database of over 2500 single studies from these meta-analyses was launched in Autumn 2021. This change was motivated by increased interest in evidence-use in education, and a desire to engage in more rigorous synthesis of primary studies. The article presents the rationale for these changes, outlines the methods adopted to populate and analyse the Toolkit database and presents results from this analysis. Findings indicate that although the broad picture of the relative benefits of the different approaches is similar, a more fine-grained analysis is possible. This deeper synthesis can provide more specific guidance about what has been successful in the different areas of the Toolkit in research studies and offers opportunities for further refinement and improvement. This increased specificity, however, comes at the cost of greater complexity in the findings and the implications for policy and practice, and it increases the challenge of ensuring findings are both accurate and accessible. A final section reflects on the challenges of summarising evidence from research to inform decision-making in education.