What Feedback Actually Does: The Evidence Behind the Advice

Most workplace advice treats feedback as good in itself: give more of it, more often, and people improve. The research literature starts from a colder position. As the organisational psychologists Gary Latham and Edwin Locke put it in 1991, feedback "is only information, that is, data, and as such has no necessary consequences at all."

Definition: Feedback is information about the gap between how someone performed and a standard. Whether it changes anything depends on what the person does with that information, which depends on where the feedback directs their attention.

That framing explains the pattern this lesson walks through. The same comment can help one person, do nothing for a second, and make a third perform worse. The useful question is never "should I give feedback?" but "what will this particular feedback make the person think about next?" The rest of this course builds on that question: this lesson covers the evidence, the second covers giving feedback, and the third covers receiving it and building a team where it flows.

In 1996 Avraham Kluger and Angelo DeNisi published the largest review of feedback research to that date, in Psychological Bulletin. They pooled 607 effect sizes covering 23,663 observations from studies where people received feedback on a task and their performance was measured.

The average result was positive: feedback improved performance by $d = 0.41$, meaning the average person who received feedback outperformed about two thirds of those who did not.

Key idea: An average effect is not a guarantee. Any feedback you give comes from a distribution in which a substantial share of cases backfire.

Kluger and DeNisi proposed Feedback Intervention Theory to explain the split. Its central assumption is that feedback shifts a person's attention between three levels:

Their moderator analyses found that feedback becomes less effective as attention moves up the hierarchy, closer to the self and away from the task. Attention is limited. Effort spent defending self-image, managing embarrassment or reading a manager's opinion of you is effort not spent on the work.

Note what the theory does not say: it does not say negative feedback is harmful and positive feedback is safe. Sign mattered less than focus.

What happens next depends on which level the comment lands on.

Education research reached a compatible conclusion from a different direction. John Hattie and Helen Timperley's 2007 review in Review of Educational Research, "The Power of Feedback", opens by stating that feedback is one of the most powerful influences on learning and achievement, "but this impact can be either positive or negative."

They sorted feedback into four levels:

1. Task: is the work correct, what is missing.
2. Process: the strategies used to produce the work.
3. Self-regulation: how the person monitors and directs their own work.
4. Self: evaluations of the person, such as "good job" or "you are clever".

They concluded that feedback at the self level is the least effective, because it carries little information about the work and draws attention away from it. The first three levels mirror Kluger and DeNisi's task focus; the fourth is their meta-task level under another name.

In practice: Two independent literatures, one on workplaces and labs, one on classrooms, agree that the object of feedback should be the work and how it was done, not the person who did it.

A 2020 meta-analysis in Frontiers in Psychology by Benedikt Wisniewski, Klaus Zierer and John Hattie pooled 435 studies with more than 61,000 participants and found a medium average effect of $d = 0.48$. Its clearest moderator was how much information the feedback carried.

Simple reinforcement or punishment, a reward or a reprimand with no explanation, had the smallest effect. Feedback that told people what was right or wrong did better. High-information feedback, covering the task, the process and sometimes self-regulation, had an effect roughly four times larger than bare reward or punishment.

Gotcha: These studies are mostly from education, where tasks and outcomes are easier to measure than in most jobs. Treat the ranking as strong evidence and the exact sizes as indicative for workplace settings.

If self-level feedback is weak, what about praise? The answer depends on what the praise is about.

The first two feel generous and cost nothing, which is why they are common. They are also the kind of feedback both reviews rank least effective, because they point at the person rather than at something repeatable in the work.

Positive feedback about the work is valuable for a reason people underrate: it tells someone which of the many things they did actually mattered. Without it, people keep what they think worked, which is often not what did.

Key idea: Make praise as specific as criticism. If you could not say what the person should repeat, the praise is mostly about your mood.

Several widely taught feedback rules come from practice or popular books rather than from research, and one was formally debunked.

Gotcha: A precise-sounding number is not evidence. The positivity ratio spread through management training for years before anyone checked the equations behind it.

Four conclusions survive across the research in this lesson, and the next lesson turns them into a method.

1. Feedback is a gamble, not a gift. On average it helps, but in Kluger and DeNisi's data over a third of interventions made performance worse.
2. Aim at the task, not the person. Both the workplace and the classroom literatures find effectiveness falls as attention moves toward the self.
3. Carry information. Feedback that says what was done, how, and what to change beats approval or disapproval alone, by a wide margin in the education data.
4. Be sceptical of formulas. Ratios and sandwiches are conventions; the positivity ratio was shown to be baseless.

Try it: Take the last piece of feedback you gave. Write down which of the three attention levels it most likely landed on: task learning, task motivation, or self. If it was the self, rewrite it so that it names one thing in the work and one thing to do next.