Background: Randomised Control Trials (RCT) involving large numbers of schools, teachers and pupils, can provide statistically significant evidence that an intervention 'works', or makes a difference to learning. However, often the quantitative data collected to illustrate the extent of impact is insufficient to illustrate exactly 'how' the intervention was enacted, what was done and 'why' it was successful. This paper collates a range of forms of data from an innovative professional training programme to indicate the nature of the promoted strategies that comprise the 'intervention' and consider how they worked in practice. Purpose: To illustrate how a mixed methods approach is required to substantiate the nature, as well as the extent of impact, of an educational intervention. Namely, Thinking Doing and Talking Science (TDTS). Sample: The project reported on here involved 42 schools in a south east county in England, UK. 21 were the 'experimental' schools and 21 were 'control' schools. Design and Methods: The project was an Educational Endowment Fund (EEF) RCT designed to assess the impact of the TDTS intervention. Results: Quantitative data showed TDTS had a statistically significant impact on the academic attainment of nine and ten-year olds, by at least three months. Various forms of qualitative data provided here offer evidential insights illustrating how and why the intervention had the impact it did on thinking and attainment. Conclusions: Designing research projects that examine both the nature and extent of impact on pupils' learning requires a mixed methods approach. This necessarily involves the statistical comparison of quantitative evidence from both the experimental and control school groupings. However, in addition to the quantitative data, qualitative evidence is required to elicit the precise nature of the intervention. This included observations during the professional development sessions, lesson transcripts, evaluative questionnaire data and interviews characterise a successful science learning intervention.