Tell me about a time you gave a teammate feedback they did not want to hear.
answer
- name the relationship and the stakes
- behavior and impact, not character
- private, timely, said to them first
- their reaction, and that you listened
- result with a number, plus follow-up
basics
~20 sTests whether you can be direct and kind at once. Answer with one specific behavior and its measurable impact, feedback delivered privately and promptly to the person themselves, and a follow-up that shows the relationship survived.
how to answer
5 beats- the relationship and what the behavior was costingTwo or three sentences: who they were to you, and the concrete cost that made silence the worse option. Keep setup and framing to roughly fifteen to twenty percent of your airtime — the commonest structural failure here is a long preamble and a ten-second ending.
- the observation you were prepared to stateSay the behavior the way you said it to them: specific, repeated, and phrased so they would agree it describes what happened. Attach the impact you observed. Never let a character word do the work a description should do.
- how, when and where you delivered itName the venue and the timing explicitly, because interviewers probe both. Private, prompt, and to them before anyone else is the defensible pattern; if you deviated, say why. State the ask you attached, so it was feedback rather than a complaint.
- their reaction and what you did with itThis beat plus the previous one carry the bulk of your airtime, around sixty percent together. Report the reaction honestly, including defensiveness, and show the part where you stopped talking. Name one thing you learned that you had not known going in.
- what changed, and where the relationship is nowClose with the result and a short reflection — about a quarter of your airtime. Give a number or an observable change, say how you followed up rather than assuming it stuck, and offer one piece of evidence that the relationship held.
your answer
5 story prompts- Pick a time you told a peer something unwelcome about their work, not a manager correcting a report.
- Choose a story where you can name one observable behavior rather than a personality trait.
- Have one number or one visible consequence ready for the result, even a small one.
- Know what you did after the conversation — a story with no follow-up has no ending.
- This can be your conflict story re-angled, if you were the one who spoke first.
draft and rehearse your own answer in a learn session
go deeper
This probes conflict tolerance and care for teammates at the same time. Interviewers want evidence that you will name a problem to the person's face rather than route around it, and that you do it with observed behavior and impact instead of character judgment. A strong answer proves the standard was raised and the working relationship survived it.
I help maintain an open-source mobile networking library, and for about a quarter I had been mentoring a newer contributor who was fast and genuinely talented. The problem was that three of his merged changes touched the background-upload path with no instrumentation behind them, and the nightly channel's crash-free session rate slid from 99.64% to 98.91% across four releases. Two other maintainers had started quietly reworking his patches after merge rather than saying anything, which I thought was worse for him than the conversation would be. I asked for a call instead of writing it in a thread, because tone does not survive text and I did not want it public. I kept to what I could observe: here are the three changes, here is the crash signature each one produced, here is what it cost in sessions, and here is what I had expected — a repro attached to the issue before merge. I said out loud why I was telling him, which was that I wanted him to end up with commit rights. Then I stopped talking. It turned out he had assumed the nightly channel was throwaway, which nobody had ever corrected. He took crash triage for the next two releases, and caught two regressions that were not his. Crash-free finished the quarter at 99.73%. He reviews on the project now, and he is the one who asks me for the blunt version.
The signal is in the observation beat: three named changes with a measured cost, no adjective about the person. The private venue and the stated intent do the kindness half. Dropping the follow-up quarter, or reporting the reaction as instant agreement, would downlevel this to a story about being right.
By then I was one of three people with release authority on the project, and the person I had to talk to was a maintainer who had been there longer than I had. His pattern was rewriting first-time contributors' patches himself and merging his own version — fast, usually better code, and it was hollowing the project out. Six of the eight people who landed a first patch that year never came back, and because his rewrites skipped the triage checklist, the beta track's crash-free session rate had drifted from 99.51% to 98.68% over seven releases. I did it one-to-one before I raised anything in the maintainers' channel, because I did not want him hearing it first as a group verdict. I brought the retention list and the crash numbers, and I said the behavior plainly: you rewrite and merge, contributors read that as their work not being wanted, and here is what it has cost. I also said what I was not claiming, which was that his code was worse, because it is not. He was defensive for about ten minutes and then asked what I wanted instead. We wrote it down: rewrites go back as a review comment or a co-authored patch, never a replacement merge, and no release ships without a triage sign-off from someone who did not write the change. I checked in with him privately after two releases. Crash-free came back to 99.69%, and five of the next nine first-time contributors sent a second patch.
Two facts lift this to senior: the feedback went sideways to someone with more tenure and no reporting line, and it ended in a written norm rather than an agreement. Naming what he was not claiming is the move that keeps candor from reading as an attack. Without the norm and the two-release check-in, this would sit at mid level.
You may have had no authority at all, and that is fine — pick feedback to a peer on shared work. What must show is that you said it to their face rather than to a third party, and that you stuck to what you actually observed.
Expect peer or feature scope: the behavior affected shared code or a shared commitment. Show the mechanics — where and when you delivered it, how you separated the behavior from the person, and that you circled back afterward instead of hoping.
Your story should carry a cost the team or production felt, and it should not end at the conversation. Name what you changed structurally so the same behavior stops recurring, and show you handled it directly before escalating.
Talk about the norm as well as the incident: feedback given to a peer of your own scope or to someone with more tenure than you, and the mechanism you left behind so the next person can give that feedback without you in the room.
saying these in an interview costs you the question
- Describing the person's character instead of one observable behavior
- Going to a manager or a group channel before going to the person
- Softening so heavily that the listener cannot tell feedback was given
- Delivering it in a public thread and calling that transparency
- Ending at the conversation with no outcome and no follow-up
- Relitigating who was right rather than what changed afterward
- How did they react in the moment?Do not sand off the reaction. Defensiveness, silence or an argument is normal and saying so is credible; what the interviewer wants is what you did next. Show that you stopped talking and listened, and that you learned something you had not known — a missing context, a constraint, a misread expectation. A reaction you describe as instant gratitude reads as invented.
- What would you do differently if you gave that feedback again?Pick one real change and be concrete: sooner, in a different venue, with evidence gathered first, or with a clearer ask attached. Avoid the humble-brag answer that you would have been even more direct. Best is a change you have actually made a habit since, because that turns a single episode into a pattern the interviewer can price.
- How is your working relationship with them now?This is the real test of the story. Answer with evidence rather than a claim — they later asked you to review their work, you two paired on something harder, they gave you hard feedback back. If the relationship did not recover, say so plainly and name what you think caused that; an honest answer here beats an implausible happy ending.
- Why didn't you take it to their manager first?Say the principle: the person hears it from you before anyone else hears it about them. Then name the conditions under which you would escalate — no change after a direct conversation and a follow-up, or conduct that is not yours to handle. Showing that you have a threshold, rather than either never escalating or escalating instinctively, is the signal.
### The prompt family This one arrives in several wordings and they are all the same question: *tell me about a time you had to give difficult feedback*, *describe a time you told a colleague something they didn't want to hear*, *tell me about an uncomfortable conversation you had with a teammate*, *how do you handle it when a peer's work isn't meeting the bar*. Prepare one story and you have answered all of them. The last wording tempts people into a general policy answer; resist it and give the episode anyway, then generalise in a closing sentence. ### What is actually being scored Interviewers are testing for a specific pair that most candidates only have one half of. Directness without warmth reads as someone who will make teammates quit; warmth without directness reads as someone who will let a problem rot for two quarters and then complain about it. The frame usually cited for the pair is radical candor — caring personally while challenging directly — with the failure mode named ruinous empathy: liking someone so much you withhold the thing that would help them. Your story has to demonstrate both halves in the same conversation, which is why a story where you were simply blunt scores badly even when the outcome was good. ### The observation, not the verdict The single most reliable structural upgrade is SBI: situation, behavior, impact. State when and where, state the behavior in words the other person would agree describes what they did, then state the impact you observed. "You're careless" is a verdict and it invites a defence of character. "The last three changes went in without a check on the crash dashboard, and we lost most of a point of crash-free sessions before anyone noticed" is an observation and it invites a fix. In the interview, telling the story this way does double duty: it shows how you talk to people, and it proves you had evidence rather than an impression. ### Venue and timing Expect to be probed on both. The defensible pattern is private, prompt, and to the person before anyone else. Public criticism — a PR thread, a group channel, a standup — converts feedback into a status event, and the recipient will spend their attention on the audience instead of on the content. Timing matters the other way: feedback saved for a review cycle is feedback the person could not act on, and interviewers hear a saved-up list as avoidance. ### Evidence that lands Rank your available evidence before you pick a story. Strongest is a behavior you observed more than once with a consequence someone else can see — a metric, a missed commitment, a contributor who stopped contributing. Next is a single incident with a clear cost. Weakest is a pattern you inferred from second-hand reports, which is a story about relaying complaints rather than giving feedback, and it will be probed hard. ### How the bar moves At junior and mid level a good answer is one clean conversation with a peer that changed something. From senior on, the conversation alone is not enough: the interviewer wants to know what you changed so the behavior does not recur, whether that is a checklist, a rotation, an explicit norm, or a written expectation. At principal scope the interesting version is feedback given upward or sideways to someone with more standing than you, delivered without organisational leverage, plus a mechanism that outlives your involvement. ### The ending Whatever you do, do not end at the conversation. The last twenty percent of your airtime is the result and the follow-up: what changed, how you know, and where the relationship stands now. Candidates who stop at "and I told them" are describing an intention, not an outcome.