Feedback has a reputation as an unqualified good. Give more of it, the thinking goes, and people improve. The evidence says otherwise, and the finding should make any manager pause.
The foundational meta-analysis on feedback (Kluger and DeNisi, 607 effect sizes, more than 23,000 observations) found a real average benefit - and that over a third of feedback interventions actually made performance worse. A later re-analysis by Wisniewski, Zierer and Hattie of 435 studies reached the same shape: a decent average effect, and huge variation, because feedback "cannot be understood as a single consistent form of treatment." What decides which way it goes isn't how much you say. It's how, and where, you point attention.
Feedback is a big part of what I teach - it's the thing students mention most in their reviews - so I've spent a long time on this question. Here's what the research actually supports.
The most reliable principle: feedback helps when it points at the task and the process ("here's what you did, here's what to try next") and misfires when it points at the self ("you're a natural," "you're so good at this"). Person-praise feels kind, but it's the weakest form. It gives someone nothing to do differently, and it can make them risk-averse - now they have a label to protect.
So take the note you were going to give - "great presence in that meeting" - and say what they did: "you paused after the client's objection instead of rushing to answer, and it made the room take you seriously." Same warmth. But now it teaches, because it names a repeatable behaviour.
Hattie and Timperley's model is a clean skeleton for any feedback: Where am I going? (what good looks like), How am I going? (what actually happened against that), and Where to next? (the most useful next step). Across the whole literature, that last one - the forward-looking step - is the most valuable and the most often missing. Plenty of feedback tells people how they did and stops there. The part that changes behaviour is the "try this next."
Most workplace feedback now travels by text - Slack, email, a comment on a doc. That's a problem, because tone doesn't survive the trip. In one study, people were confident their intended tone (serious, joking, warm) would read correctly about 90% of the time. Recipients actually got it right around 56% - barely better than a coin toss. Readers also lean negative, and that skew gets stronger when the message comes from someone senior. Which is you.
So when you write feedback, over-invest in warmth and clarity that a face-to-face chat would carry for free. State your positive intent plainly. Avoid clipped, terse lines. Then read it back once as if you were a slightly anxious version of the person receiving it - because the research says you literally cannot hear your own tone the way they will.
Burying a criticism between two compliments is one of the most common pieces of management advice, and the evidence doesn't back it. The sandwich muddies the message, and it teaches people to distrust your praise - they learn to wait for the "but." Worse, when praise is glued to the task point, people often only hear the praise, and the thing you actually wanted them to work on evaporates.
Give real, specific praise. Give a clear next step. Keep them structurally separate - different beats, different paragraphs - so the useful part survives.
While we're killing myths: there is no magic ratio of positive to negative feedback. The famous "positivity ratio" (2.9013 good comments per critical one) was mathematically debunked and formally withdrawn. Don't manage your feedback to a number.
The urge, once you're paying attention, is to list everything you noticed. Resist it. Deliberate-practice research is clear that improvement comes from focused work on one specific thing at a time. A list of six things to fix is, in practice, zero things to fix - the person doesn't know where to start, so they don't. Choose the single most useful next step and let the rest wait.
The same feedback lands differently depending on whether it feels informational or controlling - even with identical words. "You need to be more concise" is a verdict. "One thing worth playing with: could you land the headline in the first sentence?" is an invitation. Both point at the same fix. The second protects the person's sense of ownership, which is what keeps them motivated to act. Scrub "you should," "you must," "you need to" in favour of "you might," "one option," "worth a try."
Most people never get to practise giving feedback in a low-stakes setting. They do it live, on a real colleague, with real consequences, having never rehearsed. No wonder it's clumsy.
Improv is, underneath the comedy, continuous feedback practice. You give and take small course-corrections every few seconds, you learn to make your partner look good, and you build the safety that lets people hear a note without flinching. A workshop turns that into deliberate reps for a team: how to offer a correction that builds rather than bruises, and how to receive one without defending. It changes the feedback culture, not just individual technique.
A workshop that makes honest, useful feedback feel normal - and even fun. Free 30-minute discovery call to scope what your team needs.
Book a discovery callNo. It muddies the message and trains people to distrust praise. Give specific praise and a clear next step, but keep them separate.
There isn't one. The well-known ratios aren't reliable, and the most famous was withdrawn as invalid. Aim for specificity, not a number.
The work. Task and process feedback helps; praising the person is the weakest, riskiest form because it's not actionable.
Tone barely survives text, and readers skew negative - more so from a senior sender. Over-invest in warmth and clarity, and re-read as your reader would.