Embedding Formative Assessment: Evaluation report and executive summary

140 English secondary schools / 25,393 pupils (FSM subgroup n = 7,470)

Year 10-11 pupils in English secondary schools; outcome is GCSE Attainment 8, an independent national examination.

Dylan Wiliam and Siobhan Leahy's own programme tested at scale: two years of whole-school professional development via monthly Teacher Learning Community workshops plus peer observation. Pre-registered, 5/5 EEF security padlocks, 1.9% attrition. An EEF evaluation report with no DOI — the stable copy is the DERA deposit; the peer-reviewed publication of the same trial is Anders et al. 2022.

The decisive test of formative assessment: Wiliam's own programme, at scale, on independent national exams — +0.09 and marginal, i.e. 4-7x below the 1998 headline claim. At ~£1.20 per pupil per year even 0.09 would be exceptional value if real. The honest statement is that formative assessment is a cheap, small, uncertain positive rather than a transformation.

Minimal — school-level randomization with a national-register outcome.

Single large trial, unreplicated. The developer's own earlier KMOFAP quasi-experiment gave ~0.32, a ~3x hothouse-to-scale shrinkage within the same programme lineage.