The paper reports the findings of evaluative research that attempted to rigorously assess the efficacy of a feed-forward, formative assessment intervention. The aim was to improve participants' conceptions of quality, and hence improve the quality of a complex piece of summative assessment, by asking them to mark exemplars produced by former students. Feed-forward assessment has theoretical support in the literature, but empirical confirmation has been slight. Research findings were encouraging. A statistical model incorporating feed-forward was developed which accounted for a large effect in the improvement of results for the summative item. Importantly, there was improvement across student ability levels. Students, in the main, made accurate judgements about different levels of exemplar quality, although they had some difficulty discerning different pathways to high-quality products. Qualitative analysis indicated improved student conceptions of coherence and integration in the summative piece.