Feedback is supposed to be the cheap way to make people better, and on average it is. But the average hides an uncomfortable detail. In the biggest review of feedback research, feedback improved performance overall, yet in more than a third of cases it made performance worse.
of feedback interventions made performance worse than no feedback at all. Kluger and DeNisi, 1996.
When the researchers looked at why, the damage came mostly from feedback that pulled attention from the task to the self. “You’re careless.” “You’re not a team player.” “You lack presence.” Verdicts about the person.
Only 1 in 4 employees strongly agree they get valuable feedback (Gallup and Workhuman, 2024). So the skill isn’t being nicer, and it isn’t being tougher. It’s aiming.
Aim at the work, not the person
You’re careless with numbers.
In Tuesday’s report, the cost section had no comparison with budget, so the board couldn’t see we were 12% over. Next time, add a budget column and a one-line “so what”.
Why do we say the first version? Because traits are quicker to say than specifics, and because we’re all a bit prone to explaining other people by their character and missing their situation. The second version takes ten more seconds. It also gives the person something they can do on Monday, which is the whole point.
Feedback that carries real information is also far more useful than praise on its own. In one large review, high-information feedback made a far bigger difference to performance than praise and rewards did (mostly in education settings, so read across with care).
Say it sooner
The standard move is to save it for the review. It feels fairer (“I’ll have more evidence”) and it avoids an awkward moment now. But the person can’t fix what they don’t know about, and the review turns into an ambush: “Why didn’t you tell me in March?”
Give task-level feedback within 48 hours of the moment, then check back at about one, four and twelve weeks. To be clear about the evidence: feedback works as a process over months rather than a one-off event, but the 48-hour rule itself is a practical recommendation from our research libraries, not a tested finding.
Can I give you a quick thought on this morning’s client call while it’s fresh?
Skip the sandwich, and say the point first
Praise, criticism, praise. It protects the relationship, and it protects your own comfort. The evidence on it is thin either way: the one experiment on the feedback sandwich found a benefit on a maths task, and calling it proven or proven useless both overstate things.
What does have some support is clearer, not softer. In small randomised experiments in schools, a single line saying a critique came “because I have very high expectations and I know that you can reach them” sharply raised how many students revised their work. That’s a school setting, not a workplace, so it’s a lead rather than a rule. But it’s a good one.
And don’t bury the point. If you spend five minutes on “So, how are things?” before the real message arrives sideways at the end, the person spends those five minutes guessing. Say the headline in the first minute, then the reasons, then ask for their side.
A five-step shape
AHEAD
Aim. High bar. Exact example. Action next. Double-check. The Ducora feedback model, built to replace the sandwich. Five to ten minutes, within 48 hours of the event.
- Aim. Say what you’re aiming for. “What we’re aiming for is a report the board can read in ten seconds.”
- High bar, real belief. Say why you’re raising it. “I’m raising this because the bar is high on this one, and I know you can reach it.”
- Exact example. When, what happened, what it did to the work.
- Action next. One concrete next step.
- Double-check. Ask them to say back what they heard and what they’ll do. Not as a test. As a check that your message arrived as you meant it.
What we’re aiming for is a report the board can read in ten seconds. I’m raising this because the bar is high on this one, and I know you can reach it. In Tuesday’s report, the cost section had no comparison with budget, so the board couldn’t see we were 12% over. Next time, add a budget column and a one-line “so what”. Does that make sense? What will you do first?
Don’t hold it back
“They probably know already. I don’t want to knock their confidence.” This is the most common reason managers say nothing, and in our research libraries’ reading, withholding is a more common failure than harshness. The discomfort is immediate. The cost of waiting arrives later, usually in front of someone more important.
Ask yourself one question: if I were them, would I want to know?
I’d want someone to tell me this, so I’m telling you.
Praise that does something
“Great job, everyone!” doesn’t tell anyone what to repeat. Praise works better when it’s specific and tied to what it made possible, and when it arrives within a few days.
Ama, the way you rebuilt the rota so nobody did three nights in a row meant the night team had their first full weekend off in a year. Thank you.
You may have heard that you need five positive comments for every negative one. Don’t use it. The mathematical model behind that ratio was formally withdrawn.
Getting feedback back
Feedback should go both ways, and you’ll get more of it if you change the question. Ask for advice rather than feedback. “What’s one thing I could do differently in our Monday meeting?” produced more concrete, actionable suggestions than asking for feedback in a large preregistered field experiment (a strong design, but one paper, not yet replicated).
Then thank them, ask for an example so you get it right, and show what you changed within two weeks. In another field experiment, leaders who only asked for feedback didn’t make their teams feel safer a year later. Leaders who shared criticism they’d received did.
When it gets emotional
Tears or anger are normal. Don’t say “let’s not get emotional”, and don’t abandon the meeting. Name it calmly, pause, and offer a choice.
I can see this is a real blow. Take a minute, there’s no rush. Would you rather carry on now or pick this up tomorrow morning?
The evidence here is partial. Suppressing a feeling did nothing to help in research on managing emotion, while rethinking the situation and taking someone else’s perspective did. Research on naming emotions in workplace conversations specifically isn’t in our libraries yet, so this is practice informed by that research, not a finding.
What the evidence says, and how sure we are
- StrongFeedback improved performance on average, but more than a third of feedback interventions made it worse, mostly when attention moved from the task to the self (Kluger and DeNisi, 1996).
- StrongHigh-information feedback far outperformed praise and rewards (Wisniewski, Zierer and Hattie, 2020; mostly education, with a lot of variation).
- ContestedThe feedback sandwich. One small experiment on a maths task; both “it works” and “it’s useless” overstate it (Prochazka et al., 2020).
- ModerateA note saying a critique came from high expectations and belief raised revision rates in small randomised school samples (Yeager et al., 2014).
- ModerateFeedback works as a process over months, not a one-off event (London and Smither, 2002).
- EmergingGiving feedback within 48 hours. A practical recommendation from our libraries, not a tested finding.
- Strong design, one paperAsking for advice rather than feedback (Blunden et al., 2025). Not yet replicated.
- ModerateSharing criticism you’ve received, rather than just asking for feedback, raised team safety a year later in a field experiment (Coutifaris and Grant, 2022).
Frequently asked questions
How often should I give feedback?
As often as there’s something specific to say, ideally within 48 hours of the moment. Don’t schedule it for the review. Add one specific line to each one-to-one so it becomes a habit.
Is the feedback sandwich a bad idea?
The evidence is thin either way, so we wouldn’t call it proven or useless. We’d say it’s the wrong goal. The aim isn’t to cushion the message, it’s to make it clear and specific, with a visible belief that the person can reach the standard.
How do I give negative feedback without demotivating someone?
Keep it on the task, give an exact example, agree one next step, and make your belief in them explicit. Feedback damages performance when it points at the person. It helps when it points at the work.
How do I give feedback to someone I manage remotely?
Privately and, for anything tense, on video rather than in writing. Richer channels help most when intentions are unclear, and a chat thread is the worst place to land a hard message.
Sources
- Kluger and DeNisi (1996), on the effects of feedback interventions.
- Wisniewski, Zierer and Hattie (2020), on feedback versus praise and rewards.
- Prochazka et al. (2020), on the feedback sandwich.
- Yeager et al. (2014), on high-expectations feedback.
- London and Smither (2002), on feedback as a process.
- Blunden et al. (2025), on asking for advice rather than feedback.
- Coutifaris and Grant (2022), on leaders sharing criticism they’ve received.
- Webb, Miles and Sheeran (2012), on ways of managing emotion.
- Gallup and Workhuman (2024), on feedback and engagement.
