This question is the one where polite candidates accidentally disqualify themselves. They reach for a tiny stylistic disagreement, resolve it in two sentences, and inadvertently tell the interviewer they don’t push back on senior people. The real answer is the opposite: a substantive disagreement, a clean factual escalation, a clear decision, and disciplined follow-through after the call was made. That last part — what you did once the decision went against you — is where strong candidates pull away from the field.

Why interviewers ask this

Hiring managers use this question to test two things that often work against each other: independent judgment and disagree-and-commit discipline. Independent judgment is whether you’ll flag a bad decision when you see one, even when the person making it has more title than you. Disagree-and-commit is whether, once leadership has heard you and made the call, you’ll execute the decision fully — not slow-walk it, not undermine it in side channels, not say “I told you so” three months later.

Amazon’s “Have Backbone; Disagree and Commit” leadership principle is the canonical version of this idea, and Amazon-trained interviewers ask versions of this question across the industry now. The principle’s text is explicit: leaders are obligated to respectfully challenge decisions when they disagree, “even when doing so is uncomfortable or exhausting.” But the second half matters just as much: “Once a decision is determined, they commit wholly.” Jeff Bezos’s 2017 shareholder letter cited this principle as one of the four high-velocity decision habits at Amazon, and noted that disagree-and-commit isn’t a one-way street — it lets the team move fast precisely because dissent has already been heard.

The interviewer is listening for both halves. A candidate who only ever wins their disagreements sounds like they only disagree when they’re certain, which usually means they don’t surface dissent on hard calls. A candidate who only ever defers sounds like a flight risk for any judgment-heavy role. The answer they want is the one where you escalated cleanly, lost the argument, and then went and made the decision succeed anyway.

The STAR framework

The STAR beats are the same as any behavioral answer, but the weighting is unusual. You spend less time on the situation and more time on how you disagreed — the channel, the data, the tone — and on what happened after the decision.

• Situation (15-20 seconds). Briefly name the decision and what was at stake. Concrete project, concrete dollar or user impact. Don’t editorialize about your manager.
• Task (10-15 seconds). What you specifically owned, and why the disagreement landed on your desk. The interviewer needs to know you had standing to push back.
• Action (40-60 seconds). This is the longest beat. Describe (a) the substantive disagreement — what data or reasoning you had that they didn’t, (b) the channel — a 1:1, a written doc, a small-group meeting, not Slack DMs or a hallway ambush, and (c) the tone — direct, factual, not personal. Charity Majors writes that obfuscated communication does not earn trust with engineers; the same is true here. Don’t soften the disagreement to the point where the interviewer can’t tell what you actually argued.
• Result (30-45 seconds). What did your manager decide? If you won, say so plainly and credit the data, not your charm. If you lost, this is the high-bar version of the answer — describe how you committed fully, executed the plan you’d argued against, and what the outcome was. Bonus points for a learning that updated your own model, not theirs.

The whole answer should land in 90-120 seconds. Anything longer reads as score-settling.

15 sample answers

Software engineer · Won on architecture. “My tech lead wanted to ship a new feature on our legacy monolith for speed. I’d been on-call the prior quarter and had data showing two outages had originated in that exact module. I wrote a one-page doc with the incident links and a two-week extraction estimate, and asked for 20 minutes in our next 1:1. He pushed back, I pushed back harder with the on-call numbers, and we ended up extracting first. The extraction took 11 days and the feature shipped clean. He now asks me to write the doc whenever he wants to ship on the monolith.”

Product manager · Lost on prioritization, committed. “My director wanted to deprioritize an enterprise security feature for a consumer growth bet. I disagreed in writing — I had three pipeline deals worth $620k blocked on it. He heard me, weighed it, and chose growth anyway. I committed: I owned the comms to the three accounts, negotiated a six-week extension on two of them, and we lost the third. I didn’t relitigate the call in standups. Two quarters later he told me that decision was the one he’d revisit, and I think he trusted my next escalation more because I hadn’t sulked through the first one.”

Designer · Won on user testing. “My VP wanted to launch a checkout redesign without a moderated round of testing. I’d run our last three checkout studies and I had a strong prior that the new flow would confuse returning users. I asked for one week and five sessions. She gave me three. The sessions surfaced a payment-method bug that would have shipped to production. We delayed by a week, fixed it, and conversion came in slightly above forecast.”

Engineering manager · Won on hiring bar. “My skip-level wanted to extend an offer to a candidate I’d debriefed as a ‘no’ on collaboration. I asked for a 30-minute call with the full panel before the offer went out, surfaced the two specific examples that drove my vote, and we re-ran the loop with a paid working session. The candidate withdrew during the working session. We hired our second-choice four weeks later and she’s now leading the team.”

Data scientist · Lost on methodology. “My manager wanted to ship a churn model based on a six-week holdout. I argued for twelve weeks because our churn cycle was quarterly. I lost — he had a board commitment on the date. I shipped on his timeline, instrumented heavy drift monitoring, and the model held up for one quarter then degraded exactly the way I’d flagged. We rebuilt with the longer window and I never had to say I told you so because the monitoring did.”

Marketer · Won on channel mix. “My CMO wanted to put 60% of Q3 into a new influencer channel based on a competitor signal. I had attribution data showing our two highest-ROI channels were under-funded. I built a one-pager comparing payback period across channels and asked for 15 minutes before the planning meeting. We ended up at 25% on the new channel and 75% on the proven ones, and the new channel returned 0.6x — exactly what I’d been worried about.”

Sales · Lost on discount policy, committed. “My VP capped my discount authority at 15% on a deal where the buyer had a hard 22% ask. I argued the deal would close at 20%. I lost. I went back to the buyer with the 15% cap and a value-add package — extended onboarding, a roadmap seat — and the deal closed at 17%. I learned that the discount wasn’t the lever I thought it was.”

Engineering manager · Won on scope. “My director wanted my team to absorb a second product surface mid-quarter. I disagreed in a written reply to the planning doc: my team was already running at 38% on-call load and I had data on the correlation between load and attrition. He scoped the second surface to a single engineer in advisory mode instead of a full-team absorption. Attrition stayed flat.”

UX researcher · Won on sample size. “A PM wanted to make a roadmap call based on a five-person study I’d run. I told him the sample wasn’t strong enough for that scale of decision and that two participants had been atypical heavy users. He pushed back hard — the timeline was tight. I offered to run six more sessions in eight days and produce a tagged synthesis. The follow-up surfaced the opposite signal, and the roadmap went a different direction.”

Customer success · Lost on outreach cadence, committed. “My manager wanted me to cut renewal outreach from monthly to quarterly to free up time for onboarding. I had data showing the monthly touch correlated with NRR. I argued it in our 1:1, lost, and shifted to the new cadence. NRR dropped 4 points in the next quarter. I brought the data back without framing it as ‘I was right’ — just the new numbers and a proposal — and we landed on bi-monthly.”

Frontend engineer · Won on accessibility. “My PM wanted to launch a dashboard without keyboard navigation because two of our top three customers used mouse-only. I pushed back: one of the next three deals in pipeline was a government account with a hard a11y requirement. I built keyboard nav in a 4-day spike and we shipped on time. The gov deal closed eight weeks later.”

Product manager · Won on pricing. “My CEO wanted to raise our entry price 40% based on a competitor benchmark. I had win/loss data showing price wasn’t in our top five loss reasons and a cohort analysis showing the entry tier was our highest LTV. I asked for a 20-minute review and we landed on a 12% increase, segmented to the highest-ARR cohort. ARR grew, churn didn’t move.”

Engineering manager · Lost on layoff target. “My VP gave me a headcount target that I believed would cut into a critical on-call rotation. I wrote a one-page risk doc — incident frequency, MTTR, the two engineers who’d absorb the load — and asked for the cut to land on a different team. He had reasons I didn’t have visibility into and the cut stayed on my team. I executed it cleanly, restructured on-call into pairs, and we kept incident rate flat. I learned to ask earlier in the planning cycle next time.”

Senior IC · Won on technical direction. “My principal wanted us to migrate to a new database based on a vendor pitch. I’d run a benchmark against our actual workload that showed a 2.1x cost increase with no latency win. I shared the benchmark in our architecture review with the raw notebook attached. We stayed on the existing system and revisited eighteen months later when the workload shape had changed.”

Designer · Lost on visual direction, committed. “My creative director wanted a brand refresh in a direction I thought was off-strategy. I argued it in our review and lost. I shipped the refresh fully, on his vision, and didn’t half-execute it. Six months later the metrics on brand recall actually came in slightly above target. I was wrong, and the discipline of committing fully meant I learned that instead of just confirming my prior.”

What NOT to say

Closing move and practice routine

The closing move on this question is to have two stories ready: one where you won the disagreement on data, and one where you lost and committed fully. Lead with the one where you lost. It’s counter-intuitive — most candidates lead with a win — but interviewers calibrate harder on the commit story because it’s the rarer signal. If they want a second example, give them the win.

Practice this out loud against a 90-second timer. Most candidates blow past two minutes on this question because the temptation to justify the disagreement is strong. Trim the situation to one sentence. Trim the description of your manager to a single neutral clause. Spend the saved time on the substantive content of the disagreement — what data did you have, where did you say it, what did they decide — and on the post-decision behavior. If you cannot describe the post-decision behavior in two sentences, your story isn’t ready.

One last calibration check: read your answer to a friend in another function. If they can’t name (a) what specifically you disagreed about, (b) how you escalated, (c) who decided, and (d) what you did after, the answer is too soft. Sharpen it until those four facts are unmissable, and the rest of the interview gets easier — because you’ve just shown the interviewer the exact mix of backbone and discipline they came in looking for.