Most feedback makes things worse

Most workplace feedback interventions make performance worse, not better. That is not a hot take. It is what the evidence has been saying for decades. The useful question is not “how do I give more feedback.” It is how you raise the odds that this particular conversation helps.

Kluger and DeNisi’s meta-analysis is the classic citation: feedback interventions improve performance on average a little, and they make it worse about a third of the time. Unstructured feedback from people who have not been trained is even less reliable. You are playing a game where a careless move often costs more than silence.

That should change how technical leaders treat the topic. Feedback is high leverage when it works. It is also easy to spend on the wrong moment, the wrong person, and the wrong framing, then wonder why the room went quiet.

Separate growth from judgment

A lot of organisations collapse three different conversations into one annual event: how you are doing, what you will be paid, and whether you still have a future here. That mix pollutes all three.

Honest developmental feedback is easier on a random Thursday than in July when the same words might move a rating. Peer input gets softer when people know it will shape compensation. The receiver stops listening for improvement and starts defending their livelihood.

So separate the channels on purpose.

• Growth conversations are about the work and the craft. They can be frequent, specific, and low stakes.
• Performance conversations are about the formal assessment. They should be rare, documented, and unsurprising if the growth channel has been alive all year.
• Compensation should not be the only day anyone hears the truth.

If the only time you give real feedback is during rating season, you have trained people to treat every hard sentence as a threat to money. That is a structural problem, not a soft-skills problem.

Aim at the work, not the person

Whenever you can, talk about the pull request, the incident, the sprint, the release, the meeting. Not “you.” Personality feedback (“be less defensive,” “be more proactive”) is vague, hard to act on, and easy to hear as an attack on identity. Work-shaped feedback is concrete: review latency, missing tests, how a decision was framed upstairs, whether bad news arrived early enough.

Blameless language only helps if the team actually redirects blame toward the system. Saying “this is a blameless retro” and then narrating who dropped the ball is worse than saying nothing. Orient the group toward the work. When underperformance is real and persistent, escalate that to the people manager with evidence. Do not launder a performance case through casual peer critique.

Match the moment to the person

Issue 5 was about temperament. Feedback is where temperament shows up as fight-or-flight.

Some people can take a blunt note in the middle of a busy day. Others need a chosen time, a private channel, and a relationship that already has credit in the bank. David Rock’s SCARF lens is useful here: status, certainty, autonomy, relatedness, fairness. The same sentence lands differently depending on which tripwire you hit. Feedback from a senior leader about career trajectory is a status event. Feedback that rewrites someone’s plan mid-flight is an autonomy event. Feedback from someone the receiver does not respect as skilled in that domain is a fairness event.

You cannot remove those reactions. You can choose whether to walk into them blind.

A few practical rules from the coaching circle that hold up:

1. Build trust before you spend it. Early, harsh accuracy can win a battle and lose the war. If every future glance from you triggers threat, you have destroyed the channel.
2. Wait for a pattern when you can. One awkward meeting is an anecdote. Three independent mentions in a month is a pattern you can bring without it sounding like a personal campaign.
3. Do not waste a rare calm window on low-value feedback. If you only get a few clean shots a year with an anxious engineer, do not spend them on trivia.
4. Ramp intensity with the relationship. Start light. Get more direct as shared history accumulates. Familiarity builds a callus. That is good. Do not skip the callus.

Put the value up front

People go into criticism braced for harm. If the useful part of the conversation is buried at the end, they spend the whole exchange in defence and barely hear the offer.

Say early why this is worth their attention. What gets better if they take it. What you are trying to unlock. The University of Chicago writing advice applies here as well as in essays: your job is to add value to the reader, and to show where that value is as soon as you can.

Praise is not soft filler before the “real” message. The Pygmalion effect is blunt: people grow toward what gets noticed and reinforced. If Person A on the team already does the behaviour you want from Person B, praise A in earshot before you criticise B. Celebrate what good looks like close to the gap. You will often get further with less friction than a direct hit.

That does not mean never give hard feedback. It means you have more than one instrument. Use praise to shape the field. Use direct feedback when the pattern is clear and the relationship can carry it. If giving the note never makes you a little nervous, it may not be consequential enough to matter.

Small, frequent, and aimed

Ray Dalio’s “dot collector” is an extreme organisational version of a simple idea: feedback does not always need a formal write-up. A quick, specific note close to the moment (“that explanation in the design review landed”) compounds. So does a quiet correction when a public speaking habit you both know about shows up again. Aggregate beats ceremony.

What usually fails is the opposite pattern: silence all year, then a dense packet of surprises attached to pay. By then the relationship cannot absorb the load, and the content cannot be practised.

What to do with this

If you lead technically, your feedback job is not “be candid.” It is to keep a usable channel open while still telling the truth.

Separate growth from ratings. Aim at the work. Choose the moment for the person in front of you. Lead with why it helps. Reinforce the behaviour you want to see more of. Save your scarce hard conversations for patterns that matter.

If you want a mirror on the muscle, look at Mentorship and Code Review in our Big Tech benchmarks for technical leadership roles. Both are feedback systems wearing different clothes.

Resources

• Avraham Kluger & Angelo DeNisi, The Effects of Feedback Interventions on Performance (Psychological Bulletin, 1996): the meta-analysis behind “feedback often fails”
• David Rock, SCARF: a brain-based model for collaborating with and influencing others: status, certainty, autonomy, relatedness, fairness as threat/reward triggers
• Robert Rosenthal & Lenore Jacobson on the Pygmalion effect: expectations and praise shape development
• Larry McEnerney, The Craft of Writing Effectively (University of Chicago Writing Program): put the value for the reader first; the same move applies to feedback conversations

This series

This is issue 6 of a 12-part fortnightly series exploring technical leadership. Up next: goals and expectations, and how to work with corporate performance systems without letting them ruin the work.

Want the next issues in your inbox? Join the series. Email only, no account required.