Why this is not the failure question in a different coat
"Tell me about a failure" tests whether you can own a decision. This one tests something narrower and
harder to fake: whether information about yourself that you did not want can reach you and change
anything. Interviewers care because it predicts the cost of managing you. An engineer who absorbs
feedback needs correcting once; one who deflects it needs the same conversation every six months,
and eventually the team stops offering.
The material also differs. The strongest answers here are almost never about a technical gap.
"I was told my SQL was weak, so I learned window functions" is a training plan, not difficult
feedback — nobody's identity is threatened by it. Difficult feedback is about your effect on other
people, and it stings precisely because it contradicts how you understood yourself: that you talk
over people, that your reviews demoralise, that your standards read as contempt, that people agree
with you in meetings and then do something else.
Four properties of an answer that works
It is quoted, not summarised. "My manager suggested I could be more inclusive in discussions" is
a paraphrase written afterwards by the person trying to look good. "She said, in the design review
you had decided the answer before Anya finished her sentence, and everybody in the room could see it"
is what someone actually said, and it carries the discomfort. Get as close to the real words as you
can remember.
The first reaction is honest. Almost nobody accepts hard feedback gracefully in the moment, and
an answer implying you did is the least believable part of it. Saying you argued, or went quiet and
spent the evening assembling counter-examples, and then found the second and third example harder to
dismiss, is the sequence interviewers recognise from their own lives. It also demonstrates the
capability better than instant agreement would, because the interesting bit is what you did after
the defensiveness, not the absence of defensiveness.
The change is observable by a third party. "I have become more self-aware" is unfalsifiable and
therefore worthless. Something another person could point at is not.
| Vague | Observable |
|---|
| "I listen better now." | "In design reviews I now speak after everyone else has, and I write my position in the doc beforehand so I am not competing to say it first." |
| "I improved my communication." | "I stopped giving verbal decisions in reviews. Anything I decide goes in the doc within a day with the reasoning, so people can disagree in writing." |
| "I am more collaborative." | "I asked two people on the team to tell me when I had done it again. They both did, once each, in the first month." |
The loop was closed. Going back to whoever gave you the feedback, months later, and asking
whether it had improved is a small act with disproportionate signal. It costs something to ask, it
shows you treated their input as information rather than an attack, and it is the only way you know
whether the change happened or you merely intended it.
A pattern to adapt, not a script
The register is what to borrow here. The specifics are illustrative.
"About two years ago I was the staff engineer on our video pipeline, and my manager told me something in a one-to-one that I did not take well. She said that in the last three design reviews I had reached a conclusion in the first ten minutes and spent the rest of the hour explaining it, and that two people on the team had separately told her they had stopped bringing options to me because it was not worth it.
Her phrase was that I was getting the right answer and losing the team's ability to find one. My first reaction was to argue. I said the decisions had been correct, which was mostly true and completely beside the point, and I remember listing two occasions where the alternative would have cost us a quarter. She let me finish and then asked whether I could think of a review in the last year where someone had changed my mind.
I could not, and that was the sentence that landed rather than the original feedback. What I changed was mechanical, because I did not trust myself to change by intention. I started writing my own position into the design doc before the meeting so I was not fighting to say it out loud, and I made myself speak last in any review I was not running.
For decisions that were genuinely mine, I began writing them into the doc within a day with the reasoning and an explicit line saying what evidence would change them. And I asked the two engineers involved to tell me when I did it again — one of them did, about three weeks in, in front of other people, which was uncomfortable and exactly what I had asked for. Six months later I asked my manager whether it had actually changed, because I did not want to assume.
She said the reviews were slower and that two designs had gone in a direction I had initially disagreed with, one of which was better than mine. The part I still find hard is that speaking last genuinely costs time, and I have to accept a worse decision occasionally to keep the room working. I think that trade is right and I did not think so at the time."
The features doing the work: a quoted sentence, an argued-back first reaction, a mechanism rather
than a resolution, a third party who can confirm it, and a residual discomfort that is still true.
That last element is what stops the story sounding finished and therefore constructed.
The two ways candidates lose this question
The first is the flattering criticism, which is the fake failure in a new costume: I was told I take
on too much, I was told I hold too high a bar, I was told I care too much. Interviewers hear these
as an answer to a different, easier question and will usually ask for another one, at which point you
are improvising.
The second is subtler and mostly affects senior candidates: the answer where the feedback is real and
the response is entirely cognitive. You understood it, you agree with it, you have reflected on it,
and nothing in the account would look different to anyone around you. The corrective is to make
every claim about change into something with a witness — a habit someone can see, a document that now
exists, a person who was asked to check. If nobody could tell, it did not happen.
Worth having ready alongside it: one piece of feedback you considered carefully and chose not to act
on. Coachability is not compliance, and a candidate who has never rejected any feedback is either not
senior or not honest. Say what it was, why you concluded it was wrong or not worth the cost, and who
you talked to before deciding.
© 2026 Preptima. Originally published at preptima.com (opens in a new tab).