This paper reports findings from a pilot use-of-research-evidence intervention conducted with eight primary schools in and around County Durham, England, alongside an ongoing mapping study of schools’ evidence use in relation to Pupil Premium funding. In the experiment, researchers offered schools a menu of evidence-based approaches to improve attainment for disadvantaged pupils. Schools selected Learning by Questions (LbQ) for Year 4 mathematics and Student Tutoring for Year 5 English. Half of the schools were randomised to each approach for each year group. Data collection, especially of the prior attainment scores, was affected by Covid lockdown. Findings suggested that LbQ proved more feasible and was well received by teachers and pupils, though it showed little short-term impact on attainment. The tutoring intervention was more complex to implement but showed some evidence of greater improvement in the treatment group. In parallel, the paper reports preliminary findings from an ongoing analysis of schools’ Pupil Premium strategy documents. A stratified random sample of approximately 100 primary schools from each of nine regions in England was drawn. Using web crawling techniques and large language modelling, the analysis examines whether and how schools justify Pupil Premium spending with reference to research evidence. Initial results suggest that fewer than 10% of schools explicitly cite specific, directly relevant research evidence to support funding decisions. Together, the findings demonstrate the feasibility of supporting teachers’ use of research evidence through a structured intervention, while also revealing practical challenges in implementation. The pilot trial suggests potential for scaling up the intervention with a wider range of evidence-based options, a larger sample, and a longer time frame. The accompanying document analysis further situates the trial within contemporary policy and practice, highlighting the broader landscape of how evidence is invoked in school-level resource allocation.

None