What makes feedback useful

In one line: feedback can help learning a lot, but only some kinds do. The most useful feedback says where someone is going, how they’re doing, and what to try next.

What they did

This isn’t a single experiment. It’s a big review that pulls together many earlier reviews of research on feedback, mostly in schools.

John Hattie and Helen Timperley, education researchers in New Zealand, gathered 12 earlier reviews that had each combined many studies. Between them, these covered 196 studies. They asked two questions. Does feedback help people learn? And, more usefully, which kinds of feedback help, and why? From the answers they built a simple model that teachers could use.

What they found

• On average, feedback helps a lot. Across the reviews, feedback was among the strongest influences on learning that Hattie had found.
• But results vary hugely. Some feedback helped a great deal. Some did little. Some made things worse.
• Good feedback answers three questions. Where am I going? (the goal). How am I going? (progress against it). Where to next? (what to do now). The authors called the third one feed forward.
• Feedback aimed at the person helps least. Praise like “you’re so clever” carries almost no information about the task. It was the least effective kind they looked at.
• Feedback about how someone tackled the task helps most. Feedback on the method someone used, or on how they check their own work, tends to help more than a simple “right” or “wrong”.
• Most feedback in classrooms is about the task. It says right or wrong. Learners get the most from it when it also tells them what to do next.

How much should you trust it?

Strong evidence that feedback matters; less certain how well the details carry over to healthcare.

• It rests on a large body of research. But it combines very different studies, with different ages, subjects and meanings of “feedback”. That makes the average less precise than it looks.
• Nearly all the research was in schools and universities. It wasn’t tested with healthcare staff learning at work.
• The neat split into types of feedback is a simplification. In real life, timing, trust and the relationship between people all matter too.
• Hattie’s wider way of combining reviews has been criticised by other researchers (see the small print).

If you need to convince someone

In our words (a paraphrase, not a quotation): praise aimed at the person is the least useful kind of feedback. It tells people little about the task, and sometimes it can get in the way.

For a healthcare audience: “good job” at the end of a procedure is a kind word, not feedback. Both have their place, but mixing the two up is one reason so much clinical feedback is less useful than it could be.

What it means for you

Before you give feedback, check it answers three questions:

1. Does the person know the goal? Have you said clearly what good looks like?
2. Does it say what actually happened, compared with that goal? Not “good”, but what exactly was good, and why it matters.
3. Does it say what to try next time? This is the part most often left out.

It needn’t take long. To a student nurse after a cannulation: “You held the vein steady really well. Next time, try talking the patient through each step as you go.” That’s about 15 seconds, and it covers all three: an implied goal, what they did, and what to try next. The point is what they’ll do differently on the next patient.

How it appears in Teaching That Lands

Not taught directly in the three core sessions. Feedback and feedforward are parked for a future simulation-based session. The peer responses in Sessions 1 and 2 (one thing that works, one question) are a light version of the same idea.

The small print

• The 12 earlier reviews were meta-analyses: studies that combine the results of many others into one average. Together they gave nearly 7,000 effect sizes (a standard way of measuring how big a difference something makes). The average effect of feedback was 0.79, about twice the average effect of all the influences on learning in Hattie’s data.
• You may also see 0.73 quoted for feedback. That comes from Hattie’s later book Visible Learning, not this paper.
• Praise averaged an effect of just 0.14 in the paper’s summary table.
• The authors describe four levels of feedback: about the task, about the process (how the task was tackled), about self-regulation (how people monitor their own work) and about the self (the person). Process and self-regulation feedback were the most powerful; self-level the weakest.
• They also say teachers need feedback as much as learners do, for example from how learners answer questions, so they know how their teaching is landing. (Effect sizes often attached to this idea, such as 0.90, come from Hattie’s later work rather than this paper.)
• An earlier review by Kluger and DeNisi (1996) found that in about a third of the results it pooled, feedback made performance worse.
• For a balanced critique of Hattie’s methods, see Simpson (2017).