Tell me about a time you had to give someone difficult feedback.

Giving feedback is the cheapest available test of nerve

Interviewers use this question because it is unusually hard to fake and it correlates with something they care about: whether you will raise a problem while it is still small. Every team has behaviour that everyone privately discusses and nobody names — a colleague who reviews nothing, a senior engineer whose tone silences juniors in design reviews, a contributor whose estimates are consistently fictional. The question is whether you are one of the people who says it, to the person, in time.

Note that authority is not the point. This is asked of candidates with no direct reports precisely because the interesting version happens between peers, where you have no leverage and no process to hide behind. If your only example is a formal performance conversation with someone who reported to you, you have shown that you can operate a process. A peer example shows you can operate without one.

What counts as difficult

Feedback about output is easy and does not answer the question. A bug missed, a document that needed rewriting, a test suite left broken — these are corrections, and nobody finds them hard to deliver. Difficult feedback has at least one of three properties: the recipient does not believe the problem exists, they outrank you or are more experienced than you, or the subject is how they affect other people rather than what they produce.

The interpersonal kind is the strongest choice because it is the kind most people avoid. Somebody dominating discussions, dismissing juniors, going quiet after a decision went against them, taking credit ambiguously, or being unreliable in ways the team has silently started working around. Say what the effect on the work was, because that is what gives you standing to raise it at all. Feedback with no consequence attached is a complaint about style, and a good interviewer will treat it that way.

The words are the answer

Most candidates describe the intent of the conversation and never the content. "I explained that his tone in reviews was demoralising the team and that he needed to be more mindful" is a summary of a conversation that may or may not have happened. What an interviewer wants is the opening sentence, because that is the moment where nerve either holds or does not.

Softened until it disappears	What the person can act on
"I said maybe he could be a bit more mindful in reviews."	"I said that in the last three reviews he had rewritten the junior's approach in the comments before asking why she chose it, and that she had stopped proposing designs."
"I told her the estimates were a bit optimistic."	"I said that the last four estimates had come in at roughly double, that I was now privately adding to them before reporting, and that I was not comfortable doing that."
"I mentioned that we needed him to be more available."	"I said that our releases had waited on his review five times last month, and asked whether the ownership should sit with him at all."

The right-hand column shares a shape: specific instances, a stated effect, and no adjectives about the person. That is not a template so much as a consequence of doing the thing properly, and it is instantly recognisable to anyone who has had to do it.

Where these answers quietly fail

The invisible failure is the vanishing message. A candidate opens with reassurance, buries one hedged clause in the middle, closes with more reassurance, and leaves believing the conversation happened. The recipient leaves believing it went well. Nothing changes, and the candidate tells this story in interviews as a success. The tell that gives it away is a story in which the other person is not surprised, not defensive, and needs no time to absorb anything. Real feedback about behaviour almost always produces a reaction, and an answer with no reaction in it invites the interviewer to conclude that the message never landed.

The second failure is routing round the person. Raising it with their manager first, or in a retrospective as a general observation about the team, or in a document, all count as having avoided the conversation. There are situations where escalation is right — conduct, safety, anything approaching harassment — and saying so is good judgement rather than avoidance. But if the honest reason was that you did not want the awkwardness, the interviewer will hear it, and a candidate who admits that plainly and says what they would do now recovers far more ground than one who dresses it up.

What happened next

Finish with the outcome and resist improving it. The credible versions are partial: the behaviour changed in reviews but not in planning; it changed for six weeks and drifted back, so you raised it again; they were annoyed for a fortnight and the relationship survived; they disagreed, you accepted that you might be wrong, and you agreed to check with a third person. A candidate who says the person thanked them and changed permanently is describing the rarest possible outcome, and the follow-up questions are designed to test exactly that claim.

Say something about the delivery you would change, too. The most common honest regret is timing — that the behaviour had been happening for two months and the feedback would have been easier and kinder at week one. Volunteering that says more about your judgement than any part of the story itself.

The score lives in the sentence you actually said. If you cannot repeat it, the interviewer will assume the conversation was gentler than the version you are telling them.