Evidence suggests that adapting teaching responsively to pupil assessment can be effective in improving students' learning. However, existing studies tend to be small-scale, leaving unanswered the question of how such formative assessment can operate when embedded as standard practice. In this study, we present the results of a randomized trial conducted in 140 English secondary schools. The intervention uses light-touch training and support, with most of the work done by teacher-led teaching and learning communities within schools. It is, therefore, well-suited to widespread adoption. In our pre-registered primary analysis, we estimate an effect size of 0.09 on general academic attainment in national, externally assessed examinations. Sensitivity analysis, excluding schools participating in a similar program at baseline, and complier analysis both suggest a larger effect size of 0.11. These results are encouraging for this approach to improving the implementation of formative assessment and, hence, academic attainment. Our findings also suggest that the intervention may help to narrow the gap between high and low prior attainment pupils, although not the gap between those from disadvantaged backgrounds and the rest of the cohort.