The Art of Feedback That Actually Changes Student Work

Feedback is consistently rated as one of the highest-impact interventions in education — but only when it's done well. Research shows that poorly delivered feedback can actually have a negative effect on learning. The difference between effective and ineffective feedback is not effort; it's specificity, timing and content.

The meta-analytic evidence

A 2020 meta-analysis by Wisniewski, Zierer and Hattie — analysing 435 studies and more than 61,000 participants — found an overall effect size of d = 0.48 for feedback on student learning, placing it firmly in the 'above average impact' category. Critically, the analysis found significant heterogeneity: feedback's impact varies enormously depending on information content. Feedback that tells students what to do next outperforms feedback that merely evaluates performance. Grade-only feedback had among the lowest effects in the analysis.

Timing matters as much as content

Hattie and Timperley's influential 2007 model — consistently supported in subsequent research — identifies three types of feedback: feed-back (what you achieved), feed-up (what the goal is) and feed-forward (what to do next). Feed-forward has the highest impact. The 2020 meta-analysis confirmed this: feedback that includes actionable next steps outperforms evaluative commentary alone. Low-stakes, fast-feedback loops (exit tickets, mini-whiteboards, immediate verbal response) are often more powerful than detailed written comments returned days later.

Teaching students to assess themselves

The EEF's Teaching and Learning Toolkit rates metacognition and self-regulation at d = 0.60 — one of the highest available effects. Teaching students to assess their own work accurately is one of the most transferable skills in education. Students who can identify where their work meets and doesn't meet a standard develop genuine metacognition. Exit tickets give you the data you need to adjust your next lesson while also building students' self-assessment capacity.