How to Use Feedback So It Doesn't Make You Worse

"Send me your honest thoughts." You attach the draft, hit send, and brace yourself. Whatever comes back — praise, criticism, a shrug — you already know what you're supposed to do with it: take it on board, because feedback is how people get better. Your manager believes it. Your coach believes it. The entire performance-review industry is built on it. But almost nobody talks about how to use feedback — and the research is uncomfortably clear that using it wrong doesn't just waste a conversation. It can leave you performing worse than if you'd never asked.

That's not a contrarian take. It's what two organizational psychologists found in 1996 when they stopped asking whether feedback works and started counting how often it does — and how often it backfires. Their answer, plus two other research programs that attacked the same question from different angles, adds up to something more useful than "be open to criticism." It adds up to instructions.

Over a Third of the Time, Feedback Backfired

Avraham Kluger and Angelo DeNisi set out to review everything psychology knew about feedback interventions — deliberate attempts to improve performance by telling people how they were doing. For their meta-analysis in Psychological Bulletin, they considered roughly 3,000 papers. Only 131 — about 5% — met their quality bar. Those 131 studies yielded 607 effect sizes across 23,663 observations, drawn from 12,652 participants.

The average looked reassuring. Feedback improved performance, with a moderate effect (d = .41). If they had stopped at the average, the conventional wisdom would have survived untouched.

They didn't stop at the average. The distribution underneath it was the real finding: over 38% of the effects were negative. In over a third of cases, people who received feedback ended up performing worse than people who received none. And it wasn't an artifact of a few bad studies — even after they removed outliers and suspect studies, 32% of effects were still negative.

Sit with that for a second. The intervention every workplace schedules, every coach administers, and every ambitious person is told to chase made performance worse in over a third of its controlled tests. The interesting question stopped being whether to get feedback. It became: what separates the feedback that helps from the feedback that hurts?

Where the Feedback Points Your Attention

Kluger and DeNisi's explanation starts from a simple observation: feedback doesn't act on your skills. It acts on your attention. Every comment about your work points somewhere — at the details of the task, at the strategy you used, or at you, the person — and where it points determines what happens next.

When feedback keeps your attention on the task, you see a gap — between what you produced and what the standard requires — and gaps invite closing. When it drags attention up to the self, a different machinery switches on: defending your ego, replaying the comment, wondering what they think of you now. None of that effort touches the work.

The numbers track this. Feedback containing praise was associated with an average effect of d = .09 — versus .34 for feedback without praise. Read that carefully: praise didn't hurt performance. The effect stayed a hair above zero. But it nearly erased the benefit. On Kluger and DeNisi's account, a compliment pulls attention toward the self — am I talented, do they rate me — and away from the gap, and the improvement quietly disappears. Feedback designed to discourage the recipient went further, pushing performance below no-feedback controls (d = -.14). And feedback delivered under high threat to self-esteem was associated with an effect of just d = .08, versus .47 when the threat was low. Criticism aimed at the person doesn't toughen anyone up. It relocates their attention to the one place where feedback stops working.

Two honesty notes. These comparisons run across studies rather than within head-to-head experiments, so "associated with" is the accurate verb. And Kluger and DeNisi labeled their explanation — feedback intervention theory — explicitly "preliminary," with the data offering "partial support." The pattern, though, is consistent enough to act on.

What the Feedback That Helped Had in Common

Flip the analysis around and look at the feedback that worked, and three features stand out — all of them ways of pinning attention to the task.

The first row is the difference between a verdict and a correction. "This isn't working" tells you where you stand; "here's what working would look like" tells you what to do. A verdict leaves your attention free to drift up the hierarchy. A correction gives it a job.

The second row carries the biggest with-versus-without contrast of the three. Feedback framed as improvement against your own past attempts — a feature the paper labels velocity — was associated with d = .55, against .28 without it. Notice the reference point: not how you rank against other people, but whether you beat your own last attempt. If you track your goals, this is exactly the comparison your tracker data already contains — yesterday's version of you, on the record, waiting to be beaten.

The third row connects feedback to the rest of goal psychology. Feedback paired with goal setting was associated with d = .51, versus .30 on its own. Feedback tells you where you are; a specific, challenging goal tells you where you're going. Without a goal attached, even accurate feedback is a data point with nowhere to land.

Three Questions Feedback Has to Answer

A decade after Kluger and DeNisi, education researchers John Hattie and Helen Timperley reached a strikingly similar destination from a different starting point. In "The Power of Feedback", published in Review of Educational Research in 2007, they argue that effective feedback answers three questions: Where am I going? How am I going? Where to next? They call these feed up, feed back, and feed forward. Everyday feedback — a grade, a thumbs-up, a "needs work" — answers the middle question at best. And "where to next," the question that actually produces improvement, is precisely the one a verdict never touches.

Hattie and Timperley report an average feedback effect of 0.79 across 12 meta-analyses — 196 studies and 6,972 effect sizes — which they describe as twice the average effect of schooling (0.40). They are equally clear that the average comes with considerable variability: feedback in the form of cues averaged around 1.10, while praise averaged 0.14 and teacher praise 0.12. (Averaging across such different meta-analyses is methodologically contested, so treat 0.79 as their headline estimate rather than a settled constant.) The variability is the message: the same intervention, aimed differently, produces wildly different results.

Their framework also names four levels feedback can target: the task ("this opening buries your strongest point"), the process ("you're editing while you draft — separate the two"), self-regulation ("how will you check the next version before sending it?"), and the self ("you're a talented writer"). Hattie and Timperley argue that self-level feedback — praise of the person — is the least effective of the four. Different field, different methods, and the conclusion rhymes with the meta-analysis a decade earlier: the closer feedback gets to you-as-a-person, the less work it does.

Tell Me What I Did Wrong

So far the instructions look universal: point feedback at the task, skip the praise, attach a goal. But there's an asymmetry hiding in the act of asking for feedback — what you should request depends on how far into the journey you are.

In 2012, Stacey Finkelstein and Ayelet Fishbach published a series of five studies in the Journal of Consumer Research under a title that gives the finding away: "Tell Me What I Did Wrong." Across the five studies, novices sought and responded to positive feedback, and experts sought and responded to negative feedback.

In one study, they asked members of environmental organizations — the experts in this context — and novices whether they wanted to hear about actions they were performing effectively or ineffectively. 92% of the organization members chose to hear about their ineffective actions. Among novices, 74% did.

The same flip showed up in language learning: advanced students of French preferred an instructor who emphasized their mistakes over one who emphasized their strengths, rating the mistakes-focused instructor 5.45 versus 4.25 — consistent with the wider pattern in which novices gravitated toward encouragement.

Why the flip? In their studies, positive feedback increased novices' commitment, and negative feedback increased experts' sense of insufficient progress. Early in any pursuit, the live question is "is this for me — can I actually do this?" Encouragement answers it, and answering it keeps you in the game. Deep into a pursuit, commitment is no longer in doubt; the live question becomes "am I moving fast enough?" — and only information about the gap can answer that.

One caution before you reorganize your feedback life around this. These studies measured which feedback people sought and how it affected their motivation — not whether it improved their subsequent performance. And "experts" here means advanced students and environmental-organization members, not elite professionals. The claim is not that criticism makes experts perform better. The claim is that the feedback you'll actually seek out, absorb, and act on changes as you go deeper — and that's exactly the kind of self-knowledge worth using.

How to Use Feedback on Your Next Goal

Put the three research programs together and "be open to feedback" turns into something you can actually execute.

Change the question you ask. "What do you think?" requests a verdict, and verdicts point at the self. Ask instead: "What's the one thing I should change before the next attempt?" That single question points at the task, requests the correct solution rather than an error flag, and presumes a next attempt — three of the strongest features in Kluger and DeNisi's data, packed into one sentence.

Show your last attempt alongside the current one. Velocity — improvement against your own past — was the standout feature above, so engineer it into the conversation: "Here's the previous version; here's this one. Am I moving in the right direction?" You've converted a judge into a progress meter. Your own goal tracker does the same job between conversations — the chart of past-you is a comparison that always points at the task.

Leave with a goal, not a feeling. Feedback paired with goal setting was associated with d = .51 versus .30 without. So the end of any feedback conversation is one sentence: "The target for next time is X." If you can't fill in X, you received a verdict, not feedback — go back and ask the first question again.

Match the ask to your stage. Just starting out? Ask "what's working — what should I keep doing?" That isn't fishing for compliments; task-level information about what's working feeds the commitment that novices in Finkelstein and Fishbach's studies ran on. Months or years in? Ask explicitly for the gap: don't tell me what's good, tell me what I did wrong. The same rule applies in reverse when you're the one giving feedback — encourage the beginner's next step, give the veteran the unvarnished gap.

Shorten the loop when progress goes invisible. The middle of a long goal is where external results dry up and motivation sags. That's when velocity-style feedback matters most, because progress against your own last attempt may be the only progress that's visible at all. Don't wait for the quarterly review or the finished product — ask after each attempt, while the next one can still benefit.

Aim It at the Gap

That message you were about to send — "send me your honest thoughts" — rewrite it before it goes out. Try: "Here's my last attempt and here's this one. What's the one thing I should change before the next?" One sentence, and it quietly encodes most of what 607 effect sizes had to teach: it points at the task, asks for the correction instead of the verdict, frames your work as a trajectory, and commits you to trying again.

Feedback was never the unconditional good the advice industry made it out to be. In Kluger and DeNisi's data it backfired over a third of the time — not because the information was wrong, but because of where it sent people's attention. The people who improve aren't the ones who collect the most opinions about themselves. They're the ones whose feedback keeps them staring at the gap between this attempt and the next one.

So do one thing today: take the most recent piece of feedback you received — the manager's comment, the critique that stung, the compliment that felt great and changed nothing — and extract a single task-level change from it. Write that change where you'll see it when the next attempt begins. That's how to use feedback: not as a mirror to check how you look, but as a map of the ground between you and better.