Tell me about a time you disagreed with your manager.
answer
- name the call and the stakes
- why this one was worth raising
- the venue: a one-to-one, not an ambush
- their reasoning, stated back accurately
- decision, result, relationship intact
basics
~10 sTests backbone without ego. Pick one disagreement that genuinely mattered, show you raised it privately with evidence rather than opinion, and land on the decision, the outcome, and a working relationship that survived.
how to answer
5 beats- the decision on the table and what was at stakeTwo or three sentences, no more — this and the next beat together should take fifteen to twenty percent of your airtime. Name the decision your manager was making and the concrete consequence you were worried about, in terms an outsider can weigh.
- why this one was worth raisingSay out loud why you spent the capital here and not elsewhere. Referencing something you let go at the same time is a cheap, powerful signal that you pick battles rather than contest everything.
- their reasoning, stated back before you arguedGive the strongest version of your manager's position, including anything you could not see from your seat. Interviewers hear this as proof you were trying to reach the right answer rather than to win.
- how and where you made your caseThis and the previous beat are the bulk of the answer, roughly sixty percent. Name the venue — a one-to-one, a working session before the announcement — and the evidence you brought: a number, a spike, a trial. Say what you proposed, and prefer a bounded experiment over a full reversal.
- the decision, the result, and where the relationship landedClose the loop in the last twenty to twenty-five percent. State who decided what, what actually happened afterwards with one number if you have it, and one line on the relationship. Do not end on being right.
your answer
5 story prompts- Pick a disagreement with a manager or tech lead where you had evidence, not just a preference.
- Choose one from the last two years where you know how the decision actually turned out.
- Write your manager's strongest reasoning down first, before you draft your own side.
- Put one number in the result: backlog size, defects, hours, incidents, or cost.
- Have a second, smaller disagreement ready that you deliberately chose not to raise.
draft and rehearse your own answer in a learn session
go deeper
This probes backbone and judgment at the same time: whether you can voice dissent upward without capitulating and without turning it into a status fight. Interviewers read your conflict resolution and your coachability — which battles you judge worth a manager's attention, which venue you chose, and whether evidence or seniority carried the room. A strong answer ends with the working relationship intact.
I was the most junior engineer on a nine-person managed-security team at an agency, looking after patching for two client estates. A rollout took a client's login proxy offline for forty-three minutes inside a maintenance window, and my manager's response was to make every host patch manually approved from then on. I thought that would quietly grow the backlog, which was the thing the client had hired us to shrink. I didn't say that in the postmortem. That evening I pulled our ticket export and checked how the manual process had actually performed before we automated it: criticals had sat open a median of nineteen days, against four on the automated path. Then I asked for fifteen minutes in our next one-to-one. I opened by repeating his reasoning back to him — the client had been down, and he was the one who had taken that call. Then I put the two numbers next to each other and proposed something much smaller than a policy change: keep manual approval everywhere, but let me run a canary group of eleven low-risk hosts on my own account for two sprints, so we would be arguing from current data instead of my guess. He agreed to the trial. Unpatched criticals on that account went from sixty-three to seventeen with no repeat outage, and the canary group became how we staged rollouts after that.
The signal is venue and size: raised in a one-to-one rather than the postmortem, the manager's reasoning restated first, and the ask was a bounded trial instead of a reversal. Losing the two comparison numbers, or using the postmortem to score a point, would downlevel this immediately.
I owned the remediation pipeline across three client accounts. A kernel update rebooted a client's payment gateway in the middle of a weekday, and the account stayed hot for a week afterwards. My tech lead's proposal was a full stop on automated patching everywhere until the quarter closed. I agreed with the half I actually agreed with: our change windows weren't being enforced, so that reboot could have landed at any hour. What I couldn't accept was the blast radius. On the projection I ran, criticals across those accounts would have climbed past ninety before the freeze lifted. I asked for a working session before he took the plan to the client, and I brought three things. The projection. A one-sprint spike showing rollback-tested rings behind an enforced window. And the honest cost of my own option — about four engineer-days to build, plus a residual risk I couldn't drive to zero and said so. I also told him that if he still wanted the freeze after seeing all that, I would go and build the freeze properly rather than build it slowly. He narrowed it: payment tier frozen for eleven days, everything else moved onto rings. Criticals peaked at twenty-four rather than the ninety I'd projected, and the enforced windows outlived the freeze. He still routes work my way, which mattered to me more than the argument did.
Level shows in scope and in honesty about cost: the valid half of the objection is conceded, the speaker's own option is priced including residual risk, and he pre-commits to building the thing he opposed. Strip the cost sentence or the outcome number and it reads as a complaint with a slide deck.
Your scope is your own task, and that is fine. The signal at this rung is that you asked before you assumed, restated your manager's reasoning accurately, raised it once in a one-to-one, and executed cleanly whichever way it went.
Show the mechanics of building a case: the numbers you pulled, the spike or bake-off you ran, the venue and timing you picked, and what you were willing to trade. Naming the cost of your own proposal is what separates this from complaining with slides.
The disagreement should carry real risk to a team or to production, and you should price both paths out loud. Follow it with what you changed afterwards so the next call of that kind was decided on instrumentation rather than on who spoke last.
Talk about changing how decisions of this shape get made across the org: a written risk-acceptance record, a standing review with agreed criteria, clear ownership of the call. Being able to name a disagreement you deliberately did not pick is a strong principal signal.
saying these in an interview costs you the question
- Choosing a disagreement so trivial it demonstrates no backbone at all
- Badmouthing the manager or the reporting line instead of describing a decision
- Ambushing them in a team meeting and calling that courage
- Opinion only: no numbers, no spike, no cost named on either option
- Ending on who turned out to be right rather than on what actually shipped
- Never conceding anything, or folding instantly with no case made
- How did your manager react in the moment?Describe the reaction honestly, including if it was cool at first. Interviewers are testing whether you read the room and adjusted, not whether you were applauded. Say what you did with the reaction: waited, brought better evidence, or accepted the call. Avoid painting them as unreasonable; a manager who pushed back hard and then moved is a better story than one who instantly agreed.
- What did you do once the decision was made?This is the real test of the story. Say concretely what commitment looked like: what you told the team, what you built, what you stopped arguing about. If you kept a way to revisit it, name the trigger or date. Do not hint that you carried on privately disagreeing while doing the work badly.
- Was there a disagreement you decided not to raise?Have one ready. It shows you pick battles rather than contest everything. Name the thing, the reason it was not worth the cost, and what you would have needed to see to change that judgment. A candidate who claims they always speak up on everything reads as exhausting to work with.
- How would you have handled it if you had no data to bring?Answer with what you would do to get some cheaply: a timeboxed spike, a small trial, a query against existing tickets or logs. Where measurement genuinely is not possible, say you would frame it as a risk with a named consequence and ask what evidence would change their mind, rather than trading assertions.
### What this prompt is really measuring The interviewer is running two checks at once, and a weak answer usually passes one while failing the other. The first is backbone: will you say the uncomfortable thing to someone who writes your review? The second is judgment: do you know which disagreements are worth spending a manager's attention on, and how to raise one without turning it into a status contest. A candidate who never disagrees reads as passive; a candidate who disagrees with everything reads as expensive. The hire is the person who does it rarely, well, and in private. ### Common wordings you should be ready for The same family arrives as "Tell me about a time you disagreed with your manager", "Describe a time you pushed back on a decision from your tech lead", "When have you challenged someone more senior than you?", and "How do you handle it when your manager is wrong about something technical?". The last one is phrased as a policy question but is answered best with a story anyway: state your general rule in one sentence, then immediately ground it in a concrete case. Some interviewers add a constraint mid-answer, such as "and what if they had refused?" — that is a probe of your commit behaviour, not an invitation to escalate the argument. ### Weak versus strong, on the same underlying story A weak version sounds like: "My manager wanted to do it a slower way, I explained why that was wrong, and eventually he came round." Nothing in that is checkable. There is no stake, no evidence, no cost, and the resolution is that the speaker was right. A strong version of the identical episode names what was at risk, states the manager's reasoning in its strongest form before disagreeing with it, brings one piece of evidence that was not opinion, proposes something smaller than a full reversal, and ends with a measured outcome plus the state of the relationship. That last clause matters more than candidates expect — interviewers are hiring someone their own managers will have to work with. ### Choosing the episode Pick a disagreement where three things are true: the stake was real, you had or could get evidence, and you know how it turned out. Avoid disagreements about people, promotions, or someone else's performance; those pull the conversation somewhere you cannot win. Technical direction, sequencing, risk acceptance and process changes are all good ground. It is entirely acceptable to pick a case where you lost the argument and the manager turned out to be right, provided you can show the case you made was substantive and your commitment afterwards was genuine. ### Venue and timing carry a lot of the signal Say explicitly where you raised it. A one-to-one, a direct message asking for fifteen minutes, or a working session before the decision was announced all read as professional. Raising it for the first time in a group meeting, in front of the manager's own manager, or in a channel where the client can see, reads as either naive or political. If you did raise something publicly, have a reason: some decisions genuinely must be challenged in the room where they are being made, and saying so deliberately is stronger than pretending you always take things offline. ### How the bar moves Early-career, the interviewer mostly wants evidence that you did not silently comply and did not sulk. Mid-level, they want the mechanics: how you built a case and what it cost. Senior and above, they are listening for whether you priced both options honestly, whether you left the organisation with something durable — a measurement, a review trigger, an agreed criterion — and whether you can describe a disagreement you chose to drop. Across every level, the answer should be shorter than you think: about fifteen to twenty percent of your airtime on setup, the bulk on what you actually did, and a firm final twenty percent on result and reflection.