Tell me about a time you got feedback you disagreed with.
answer
- name the feedback in their words
- steelman it before you rebut
- the kernel that was true
- evidence, not opinion, in the reply
- the decision you carried out
basics
~20 sTests whether you can hear criticism you believe is wrong without getting defensive. Steel-man it, name the part that was true, push back with evidence instead of opinion, and end on a decision you actually carried out.
how to answer
6 beats- the feedback, in their words, and who gave itQuote it as neutrally as it was said, and name the giver by role rather than by name. Keep this setup plus the stakes to roughly fifteen to twenty percent of your airtime — the interviewer needs only enough context to judge the disagreement.
- why you disagreed, and the strongest version of their caseSay what you believed and why in one or two sentences, then immediately give the best argument for their side — what would have to be true for them to be right. Doing the steelman before the rebuttal is the move that stops this sounding like a vindication story.
- what you did to test it before arguingThis and the next beat are the bulk of the answer, around sixty percent together. Be concrete: you asked which specific piece of work prompted it, you pulled the record, you checked with someone else who saw the same work. Show that you tried to disprove yourself first.
- how you brought the disagreement back, with evidence and a concessionDescribe the actual conversation: the evidence you put on the table, and the part of the critique you conceded. Naming the kernel that was true is what earns you the right to hold the rest, and it is the beat weak answers skip entirely.
- the outcome and the decision you carried outLand the result in about twenty to twenty-five percent of the airtime, with one number or verifiable change. Say plainly which of the three endings happened: they moved, you moved, or you disagreed and committed and then executed it properly.
- what you still hold and what you changedClose with one honest line on the part you would still argue and one on what you do differently now. Ending on pure vindication or pure surrender both read badly; the credible ending has a piece of each.
your answer
5 story prompts- Pick feedback you genuinely believed was wrong at the time, ideally from the last two years.
- Write the giver's strongest argument in one sentence before you write a word of your rebuttal.
- Name the one part you conceded — an answer with no concession sounds rehearsed.
- Find the record you used instead of an opinion: a ticket trail, a log, a number.
- This can be the same episode as your tough-feedback story, re-angled onto the part you contested.
draft and rehearse your own answer in a learn session
go deeper
This probes coachability under genuine disagreement, which is where it actually gets tested. The interviewer wants to see you separate the substance of a critique from the person delivering it, test the claim before rejecting it, and argue with evidence rather than status. A strong answer proves you can hold a position, drop one when the record moves, and commit to an outcome you did not choose.
I was about a year into my first job, at a seventeen-person security consultancy, and I ran the quarterly access reviews for one of our clients. In a check-in, my team lead told me my findings write-ups were bloated and that nobody was reading them, and that I should cut each one to a single line. I disagreed, because the detail was exactly what let the client's platform team confirm a revocation was safe. Before I argued, I tried to make his case for him. I re-read three of my recent reports as if I had forty of them sitting in my inbox, and he was right that they were unreadable. I just thought the problem was structure rather than length. So I asked him which report he had bounced off, and what he had skipped in it. Then I went to the ticket history: of the forty-one findings I filed the previous cycle, twelve came back asking which account I meant, so the detail was not landing either. I brought him that number and proposed a middle version, with the decision and the account on line one and the evidence underneath. We ran it for a cycle. The client closed sixty-six of seventy-one flagged accounts inside the review window, against forty-four of seventy-one the cycle before, and the follow-up questions mostly stopped. I still do not think length alone was the issue, and I said so. What changed is that I stopped defending my format from memory and went and measured it first.
The signal sits in the test-before-rebut beat: he steel-mans, asks which specific report triggered it, then checks the ticket history instead of arguing from recollection. The measured result and the honest residual disagreement carry the ending. Dropping the number, or conceding nothing at all, would downlevel this.
I was running a seven-week access-review engagement for a client, leading two other consultants. Halfway through, our engagement director gave me hard feedback: I was being obstructive with the client's identity team, and my escalation had cost us goodwill. My first reaction was that it was unfair, because I had escalated in week five, after they missed three agreed handoffs. I asked him to tell me exactly what he had seen, because I wanted his version before I defended mine. He had been on the call where I listed the missed handoffs in front of the client's VP. That reframed it: he was not telling me the escalation was wrong, he was telling me I had done it in the worst possible room. I still brought the record, because I did not want the substance quietly conceded. I showed him the handoff log, three of the four promised extracts late by a median of nine days, and said I would escalate again on the same facts. Then I agreed with the part I had got wrong. We settled on a rule I have used since: the miss goes in writing to my counterpart lead first with a forty-eight-hour window, and only then into a shared forum. I also went back to the identity team's lead and apologised for the venue, not for raising it. That mattered more than anything I said upstairs. They unblocked the remaining extracts that week and we finished at one thousand two hundred forty of one thousand three hundred eighteen accounts reviewed. I would make the same call again. I would just make it twice quietly before I made it once loudly.
Separating substance from delivery is the whole move: he holds the escalation, concedes the venue, and leaves a reusable rule behind. Going back to the blocked team is the ownership beat, and the handoff log is what keeps it evidence-based rather than a capitulation dressed up as growth.
Scope is your own output. Show that you asked which specific piece of work triggered the feedback, checked the record before arguing, and made one measurable change even if you kept part of your position.
Scope is a feature or how you work with peers. Separate substance from delivery, bring evidence a peer could independently check, and close the loop with the giver rather than quietly deciding they were wrong.
Scope is your team or a production call. The interviewer wants the decision you made under the disagreement — escalate with evidence, or disagree and commit — plus the rule or check you left behind so it is not re-argued next quarter.
Scope is strategy or org direction. Show you tested the critique with people outside your reporting line, and that you changed either the direction itself or the mechanism by which disagreements of that kind get resolved.
saying these in an interview costs you the question
- Relitigating who was right instead of describing what you did about it
- Picking feedback you secretly agreed with, so nothing is actually at stake
- Framing the giver as biased, incompetent or out to get you
- No steelman — the critique is dismissed before it is ever tested
- Ending on "I was right" with no change and no decision carried out
- Impressions and feelings as evidence where a log, ticket or number existed
- What did the person who gave you that feedback say afterwards?Interviewers use this to check the relationship survived. Describe the closing conversation concretely — you went back with what you found, what you changed, and what you still saw differently. If the relationship stayed working, say so with a small proof: they kept giving you feedback, they asked you to run the next one. Never let the answer imply you simply stopped talking to them.
- What would you have done if they hadn't changed their mind?This is the disagree-and-commit probe. Say where your threshold sits — one round of evidence, then the decision belongs to whoever owns it — and describe committing without sandbagging: you execute the version you argued against and you do not narrate your objection to the rest of the team. Adding what you would have measured to reopen it later shows judgment rather than resignation.
- Which part of that feedback do you still think was wrong?Answer it straight — the trap is pretending you were fully converted, which reads as performance. Name the specific claim you still reject and the reason, in one or two sentences, without contempt for the person who made it. Pair it with the part you did concede, so holding a position looks like discrimination rather than stubbornness.
- How did you decide it was worth pushing back on at all?They are testing your escalate-versus-let-go threshold. Give a real filter: does the feedback change what someone else will do, does it stick to my record, is the cost of being wrong reversible. Say out loud that you let smaller things go, and give the tiny example — a candidate who contests everything is exhausting, and a candidate who contests nothing has no spine to test.
## The harder cousin This prompt is the harder cousin of "tell me about the toughest feedback you've received". There, the expected arc is listen, accept, change. Here the interviewer deliberately removes the easy ending: you thought the critique was wrong, so the answer has to show what you do when **coachability** and **conviction** pull against each other. ## Wording variants you should be ready for - "Tell me about feedback you disagreed with." - "Tell me about a performance review you thought was unfair." - "Describe a time you were criticised and you didn't think the criticism was justified." - "Tell me about feedback you decided not to act on." All four are the same family and one story serves them, but they land on different beats: the **review** variant pulls for how you handled a formal, written, permanent assessment, and the **"decided not to act on"** variant pulls hardest on your threshold for letting something go. ## The two failure modes 1. The first is the **doormat answer**, where the candidate says they disagreed and then simply did what they were told, with no test and no argument. It reads as compliance, not judgment, and it usually comes from a fear that any disagreement sounds difficult. 2. The second is the **vindication answer**, which spends its whole airtime proving the giver was wrong. Everything in it points backwards: what they misunderstood, what they hadn't seen, how it turned out exactly as the candidate predicted. Interviewers hear both often, and both fail for the same reason — neither shows a decision being made. ## What a strong answer does instead 1. It opens by **quoting the feedback** the way the giver said it, not the way it felt. 2. It **steel-mans**: one sentence on what would be true if they were right, delivered before any rebuttal, which is the single move that separates this answer from the vindication version. 3. It then goes and **tests**, which is where most of the airtime should go — asking which specific piece of work triggered it, pulling the ticket history, the log, the review comments, the data the disagreement is actually about. 4. And it comes back with two things at once: **the evidence, and the concession**. Almost every real disagreement has a kernel that was true, usually about delivery, timing or how the work landed rather than the substance. Naming that kernel out loud buys you the right to hold the rest. ## Evidence types, in rough order of strength A record that existed before the disagreement — tickets, logs, review history, a handoff trail — beats a measurement you took afterwards, which beats a colleague's opinion, which beats your own recollection. If your story's only evidence is "other people agreed with me," expect the interviewer to push on it, because that is a popularity argument in the shape of a data argument. ## The ending is where levels separate Three endings are all legitimate: - they moved, - you moved, - or nobody moved and you committed anyway. Do not manufacture the first. If your story genuinely ends in **disagree-and-commit**, say the words — you argued once with evidence, the decision belonged to someone else, you executed it properly and did not undermine it in the corridor. That ending is often stronger than winning, because it proves you can be overruled without becoming a problem. ## How the bar shifts - A **junior** is judged on whether they checked before they argued at all. - A **mid-level engineer** is judged on separating the substance of the critique from its delivery, and on closing the loop with the giver rather than privately filing them as wrong. - A **senior** is judged on the decision made under uncertainty and on what they left behind so the same fight does not recur — a written rule, a check, an agreed escalation path. - A **principal** is judged on whether they tested the critique outside their own reporting line and changed either the direction or the mechanism. ## Finally, pick the story with real stakes Feedback about something you secretly agreed with, or a disagreement that cost you nothing, gives the interviewer nothing to score. The best version is one where you were partly right, partly wrong, said both out loud, and the working relationship is still intact today.