Describe a Significant Disagreement During a Project
The question: "Describe a situation where you had a significant disagreement during a project. How did you handle it and what was the outcome?"
Also asked as: "When did you disagree about what to cut to hit a date?" · "Tell me about a time a decision went against you. What did you do next?" · "When has a disagreement put a delivery date at risk, and how did it close?" · "When did you disagree with an executive on a program's date?"
Who's asking: an Engineering Director or Senior Engineering Manager, in a behavioral round.
Executive Summary: The director wants to know what happens to a decision when you disagree with it. Does it close sooner or later because you're involved, and what do you do the morning after it closes against you? The trap is a story with no clock, where two smart people argue until everyone agrees you were right. What changes with level is who has to close the decision: you and a teammate, your project and its PM, several teams and their managers, then an executive holding the money.
What the director is really asking
The director is answering three questions:
- "Does disagreement slow this person down or speed up the decision?" They listen for the date, the day it closed, and what you did to close it.
- "Can I count on them when it closes against them?" Someone who commits only after winning is a risk on every project, so they want to hear what you did next.
- "Did they understand the other side?" If the other person's argument sounds weak in your telling, the director assumes you never heard it.
This is the question with a calendar in it: the date is what makes the disagreement hard. Friction between two people belongs to "conflict with a coworker", and you against a group's plan to "going against a team's proposal".
It maps closely to Amazon's "Have Backbone; Disagree and Commit", and loosely to "Deliver Results"; treat both as approximate. Three ladder dimensions do most of the work: conflict (who you can close a decision with), ownership (what you do once it's closed), and escalation hygiene (when you take it up, and what you bring).
Level fit at a glance
| SDE | Senior SDE | Staff | Principal | |
|---|---|---|---|---|
| Asked at this level? | Core | Core | Common | Rare |
| Story scope they expect | You and a teammate, over your piece of a release | A scope or date fight on a project you lead | Several teams sharing one date, often with no single owner | An exec-level disagreement over a program's date or scope, in money |
| The bar in one line | Disagrees respectfully, closes it quickly, knows when to involve the lead, ships the decision either way | Settles it with evidence before the date does, and escalates cleanly when it must go up | Brokers teams with competing incentives to one decision by a date, and aligns their managers | Argues in the terms the exec decides in, then changes their mind or commits visibly |
| Instant down-level | "We argued until the lead picked one" | "I was the lead, so we went with my plan" | One team against its PM, told big | Technical risk with no money in it, ending in "and I turned out to be right" |
Every strong answer names the date, the other side's best argument, who decided and when, and what you did next. Can you say the date, and the day it closed? If not, pick another story.
SDE: close it fast, and ship the decision even when it isn't yours
What meets the bar
- The disagreement has a date on it, and you can say why that date couldn't move.
- You can state your teammate's strongest argument as well as they could.
- You settled it together, or took it to your lead with both options once you'd both stopped moving.
- You own your part, including a number or assumption that turned out wrong, and you raised it early.
- You shipped the decision, and helped, even though it wasn't yours.
Model answer
"Our company's new fitness band was launching in October on a fixed date, and the companion app had to reach the store ten days before that. I was a mid-level engineer on the app's sync team. A week and a half before that cutoff, beta users started seeing duplicate workouts whenever the band reconnected, which inflated their weekly totals.
A teammate who'd worked on sync longer wanted to filter the duplicates when the app read the workout list. I wanted to fix the sync cursor so duplicates never got written. We went back and forth in a review thread for a day and a half. His argument was a good one: the filter was a day's work and couldn't damage anyone's history, while my change touched code every sync goes through, right before a launch. Mine was that his left bad rows on the server, so the weekly summary email would stay wrong.
Neither of us was moving, and the cutoff wasn't either, so I asked him for twenty minutes with our lead. I put both options on one page, with risks and days, and he checked I'd described his fairly. I said my fix would take three days. Our lead gave me those three days, with his filter ready as the fallback and a decision set for Thursday noon.
By Wednesday morning I knew my number was wrong. I'd only looked at current firmware. Older bands send local timestamps, and handling them safely was more like eleven days. I told them both that morning, and our lead called it for his filter. I reviewed it and wrote the tests for older bands.
We made the cutoff with four days to spare. On-screen duplicates dropped from about six percent of beta syncs to under half a percent. The email stayed wrong for six weeks, until a cursor fix replaced the filter, which was the trade we'd picked.
Any number I give for a fix I'm arguing for gets run against the oldest device we support first, and I say out loud what I haven't checked."
Why this lands at SDE
- "Neither of us was moving, and the cutoff wasn't either": the clock is why they went to the lead, and "he checked I'd described his fairly" shows how to take a disagreement upward.
- "The filter was a day's work and couldn't damage anyone's history": the teammate's strongest case, in the candidate's own words.
- "I told them both that morning, and our lead called it for his filter": the level beat. The candidate raised their own miss a day before the checkpoint, then did the work on someone else's fix.
- "The email stayed wrong for six weeks": the prediction came true, yet it claims no win: the filter was the right call for that date.
Follow-ups to expect
- "What did your teammate say about you afterwards?": tests how you treat people → what he actually said, including anything uncomfortable.
- "Why take it to your lead, and not settle it yourselves?": tests when to ask for help → what you'd tried, why the date made more arguing expensive.
- "You said three days and it was eleven. What did that do to how your lead hears your numbers?": tests coachability → what you now check first.
Red flags
- ❌ "I was sure, so I built my fix anyway." Going round the disagreement is the worst version of this story.
- ❌ "My lead picked one and I moved on." The action was escalation, and there's nothing of yours in it.
- ❌ A teammate who "just wanted the quick hack". A safe fix before a launch is a fair position.
Senior SDE: settle it with evidence, before the date decides for you
What meets the bar
- You lead the project, and you own the outcome and the date.
- You test your own case, and you say so to the other person when the evidence breaks it.
- You force a decision by a date, and when it goes up, you escalate cleanly: one page, both sides, a choice to make.
- You can name what each side gave up, as a visible trade with your PM.
Model answer
"Our grocery chain's holiday circular went to print on October 10th, and once it did, its double-points offer was a promise in 320 stores. I was tech lead for the loyalty team, six engineers, moving about four million members off a vendor's points system onto our own ledger. The old system couldn't run that offer.
In late August our PM wanted every store cut over by early November. I wanted to wait until January and print a simpler offer. She pointed out that the migration had already slipped past two holiday freezes, and every slip meant another year of vendor fees. My case was the dry run, where 0.4 percent of balances didn't match. I told her we'd hand 16,000 members wrong balances at Thanksgiving, and we went around on that for nearly two weeks.
Then I asked one of my engineers to trace the mismatches. Almost all were rounding on expired points, worth pennies, and a week's work to fix. I had the wrong risk. The danger was rollback: once a store earned points on the new ledger there was no way back, so a bad November Saturday meant cashiers couldn't apply points. Neither of us had been arguing about that.
I told her that same day that my January case didn't hold, and why. I proposed going by region, starting with 40 stores by October 3rd, plus a reverse sync so any region could fall back. That sync took two engineers three weeks, so her redesigned points-history screen came out of the launch. I wrote one page for our two managers, and she edited it: the date, what each of us gave up, and the go/no-go test for those 40 stores. They signed it on September 11th.
The circular printed on time, and all 320 stores were across six days before the promo. We used the fallback once, one region, for five hours. About 2,300 members had a wrong balance in the first week. The history screen shipped in February.
Ask that PM today and she'd still say I spent two of our six weeks defending the wrong number, and that the screen she gave up missed the season it was built for. I don't have a good answer to either."
Why this lands at Senior SDE
- "I had the wrong risk": the candidate tests their own number, it fails, and they say so.
- "I wrote one page for our two managers, and she edited it": the level beat. A joint page with the trades on it leaves the managers nothing to referee.
- "She'd still say I spent two of our six weeks defending the wrong number": the candidate says the unflattering part before the director can ask.
Follow-ups to expect
- "How did your manager see the two weeks you spent on the wrong risk?": tests self-awareness → what your manager said, and what you now check before arguing from a number.
- "What did you give up to hit the date?": the Senior trade-off probe → the January date, her screen, two engineers for three weeks, and who agreed.
- "If I called that PM, what would she tell me about you?": tests how you treat partners → the unflattering part first.
- "Who grew on this?": tests the people dimension → the engineer who traced the mismatches, and what they did next.
Red flags
- ❌ "I was the lead, so we went with my plan." Resolved by authority, which is an instant down-level.
- ❌ "The PM only cared about the date." The date was a printed promise and two earlier slips. Say that instead.
Quick check before moving on:
Staff: get several teams to one decision by a date
What meets the bar
- The date is shared by several teams with different managers, often with no single owner, and you can say why it fell to you.
- You got their managers to one decision by a date, without authority, and you can name the incentive behind each team's position.
- Decisions don't queue on you, and if they did at first, you say so.
- You credit the teams, and name the cost in org terms.
Model answer
"The league's season opener was a fixed date for our sports streaming service, eleven weeks out. As a staff engineer outside the playback, ads and entitlements teams, I laid their plans side by side, and they didn't fit. Playback wanted a low-latency mode, five seconds behind live, because viewers hear about a goal on their phones before they see it. Ads had sold stitched breaks that needed about ten seconds of lookahead. Entitlements needed players to stop re-checking every subscription at kickoff. Each case was sound, each team answered to a different director, and no plan covered the whole night.
So I asked the three managers for one joint cut list, decided within two weeks, with one test: what fails in front of our biggest audience of the year. I argued to cut low latency entirely. Playback's manager pushed back hard: the opener is when people try the service, and a stream that spoils the goal is what they'd remember. That was a retention argument, and I only had a risk one. On day twelve they signed it: low latency on connected TVs only, at about twelve seconds; ads dropped the personalized kickoff break it had sold; playback and ads both took the entitlements change.
Then I became the bottleneck. To keep their list consistent, I'd offered to be where every cross-team interface question came. Four weeks in, nineteen were queued on me, and ads waited nine days for a marker-format answer their lead and playback's lead could have settled in an hour. Their manager told me so, fairly. That week I made a pair of leads the owners of each interface, with the cut list as their test, and kept only the calls that changed the list. Waits fell to under a day.
The teams built it and it held: 1.4 million concurrent viewers at peak, entitlement errors under 0.1 percent. Ad fill was 91 percent against the 96 we'd sold, from a stitching bug on one device family, and make-goods ran about $210,000.
Those nine days took out our second full rehearsal, on the last preseason game, and I think that rehearsal finds the bug. So my queue likely cost that $210,000. I said that in the review, in front of all three managers."
Why this lands at Staff
- "I laid their plans side by side, and they didn't fit": the candidate found the problem, in a gap between teams.
- "I asked the three managers for one joint cut list, decided within two weeks, with one test": the level beat. A date and a criterion, with no authority over any of the three.
- "That was a retention argument, and I only had a risk one": directors trust someone who can name the argument that beat them.
- "Those nine days took out our second full rehearsal": the bottleneck is priced in org terms, and said to all three managers.
Follow-ups to expect
- "Why was this your problem to solve?": the Staff probe → you'd seen the misfit, no plan covered the night, and what you stopped doing to make room.
- "You argued to cut low latency and lost. What did you do then?": tests commitment → what you did that week to help playback's plan.
- "What did the ads manager say about you afterwards?": tests how you treat peers → the nine days, what she said, and how that relationship stands.
- "How did you get three managers to sign one list?": tests alignment → what each of them needed, and who gave up what.
Red flags
- ❌ "They couldn't agree, so I decided." You had no authority, so what did you actually do?
- ❌ Staying the single point of decision and calling it "consistency". A queue on you costs several teams.
Principal: argue in the exec's terms, then commit when it closes against you
What meets the bar
- Why it's rare: directors seldom ask a Principal this in these words. It usually comes as the executive version above, and Amazon-style loops tend to probe it at every level. Either way it tests disagree-and-commit at org level, in the terms execs decide in.
- Keep the scope honest: one program and one date is the right size.
- You made the case in the exec's terms, with a price for both outcomes, and you can give their best reason, whichever way it closed.
- If it closed against you, commitment is visible: what you stopped saying, moved, gave up.
Model answer
"Our regional bank was replacing its core ledger: about 900,000 accounts, three orgs, and a conversion that had to land on a long weekend. I was the principal engineer accountable for the technical plan. Our COO had set Labor Day. Ten weeks out, I thought reconciliation wasn't ready, and I argued for the October long weekend.
I went in with defect counts and batch timings. She heard me out, then asked what a slip cost and what a failure cost. I had neither number. That was my mistake: I'd argued engineering risk to someone deciding with money, and it cost us a week. I came back with finance: $4.8 million to slip, perhaps $60 million for a failed conversion, at odds I put near one in ten. She held Labor Day. The slip was certain, my odds were a judgment call, and a delay that late would have cost us credibility with the board and our examiners. Those are good reasons.
That day I stopped raising the date. I moved two architects onto reconciliation, gave up a quarter of my own roadmap, and wrote the go-live criteria with her risk team, so there was one agreed reason to reopen it.
We went live on the date. About 6,000 accounts needed manual correction in week one, costing roughly $1.1 million. I still think October was safer, though I have no evidence. Committing meant my leads didn't hear that from me until after the go-live review."
Why this lands at Principal
- "I'd argued engineering risk to someone deciding with money": the correction this level has to make, named as the candidate's own miss.
- "Wrote the go-live criteria with her risk team, so there was one agreed reason to reopen it": the level beat. Full commitment with an honest way back, built with the exec's own function.
- "The slip was certain, my odds were a judgment call… Those are good reasons": about $6 million expected against $4.8 million certain is close, so her objection to the odds decides it.
- "I still think October was safer, though I have no evidence": a business result ($1.1 million against $4.8 million) that doesn't settle who was right, and no pretending it does.
Follow-ups to expect
- "She overruled you. Were you right?": tests ego → what the result does and doesn't prove, and no rewriting of it in your favor.
- "When your leads asked whether you agreed, what did you say?": tests integrity → the date was the plan and you'd had your say, without lying about what you think.
- "What would have made you go back to her?": tests that commitment isn't silence → the go-live criteria, agreed in advance.
- "What mechanism outlived your involvement?": the Principal probe → the honest answer here is little. Say so, name what you'd build, and offer a story that has one.
Red flags
- ❌ "I told them the risk, and they ignored it." Contempt for the exec, and your framing was the miss.
- ❌ A story ending "and it went badly, as I'd said". Directors hear someone waiting to be proved right.
Quick check before moving on:
One story, four altitudes
| SDE | Senior SDE | Staff | Principal | |
|---|---|---|---|---|
| Situation | Duplicate workouts, app-store cutoff days away | A ledger cutover against a printed holiday circular | Three teams' plans that don't fit one season opener | A core-ledger go-live you want to move |
| Your role | Engineer on sync, arguing for your own fix | Tech lead of a six-person loyalty team | Staff engineer, no authority over the teams | Principal for the technical plan, three orgs |
| The key move | Took a stalled thread to the lead; flagged your own bad number | Dropped your own case on evidence; one page for two managers | One joint cut list by a date; gave interface decisions to pairs of leads | Came back with money, lost, stopped raising the date |
| Result | Cutoff made; on-screen duplicates 6% to under 0.5% | 320 stores across six days early; 2,300 wrong balances | 1.4M viewers, errors under 0.1%; ad fill 91% vs 96%, $210k make-goods | On date; 6,000 corrections, about $1.1M, against $4.8M to slip |
| Reflection | Test your estimate on the oldest device first | She'd still say you spent two weeks on the wrong number | Your queue cost a rehearsal, so say so | Your leads didn't hear your doubts until after the review |
Draft your answer
Step 1: Pick the right story
Run each candidate story through the test for your target level:
- SDE: Was there a cutoff, did you know when to involve your lead, and did you ship the decision either way?
- Senior SDE: Did you lead the project, change your position or force a close by a date, and trade something visible with your PM?
- Staff: Did at least two teams besides yours change their plan, can you name each one's incentive, and did it close by a date you helped set?
- Principal: Was an executive deciding, was money on the table, and can you say what changed their mind or what committing cost you?
Then check that it passes all of these:
- It's from the last two or three years, and you were central to it.
- There's a date, and you can say why it couldn't move.
- You can give the other side's strongest argument, and you'd be happy for them to hear it.
- You can say who decided, and on what day.
- There's a number in the result, and something it cost.
Don't reach for the story you won. One that closed against you is often your strongest, because it shows you can commit.
If you're a new grad. A hackathon, a course project or a lab paper counts when a fixed deadline, like judging or a submission date, forced the call, and you put your case to whoever decided. The director expects a teammate and a small piece of work, but still wants the date, their best argument and what you did once it closed: "Judging was at noon, and at 2 a.m. we still disagreed about cutting the login screen."
Aiming a level above your current title? A Senior aiming at Staff can tell the several-teams version only if other teams' managers actually changed their plans. If none clears the level, tell your best project-level story, and let the reflection show the bigger version: "Next time I'd get the register team's manager on that page in week one, because their freeze date was tied to ours." An honest Senior story with Staff-level awareness beats an inflated one that falls apart at "Who actually decided?".
Step 2: Fill the skeleton
SITUATION (15s) — Where you were, your scope, and the date that made
this matter: ______
DISAGREEMENT (20s) — What to cut, when to ship, or who decides. Your position
in one line, then the other side's STRONGEST argument: ______
ACTIONS (45s) — 2–3 moves, each "I …, because …". How you got it to a
decision, and the thing you personally got wrong: ______
CLOSE (10s) — Who decided, on what day, and which way it went: ______
COMMIT (10s) — What you did next, especially if it went against you: ______
RESULT (15s) — Did the date hold? One number, and what it cost: ______
REFLECTION (15s) — What you do differently now, or what the other person
would still say: ______
LEVEL BEAT (inside ACTIONS or COMMIT) — the line only your level could say.
SDE: the miss you raised early, or the decision you helped ship.
Senior: how you forced the close, and the trades you and your PM named.
Staff: what each team gave up, and how their managers came to sign.
Principal: the money you argued in, and what changed the exec's mind or
what committing cost you.
Aim for about two minutes, two and a half at most, with a 30-second cut ready (the date, both positions, who decided, what you did next, the number).
Step 3: Calibrate it
| If your answer says… | It reads as… | Fix |
|---|---|---|
| "We debated it for weeks until it got sorted" | No clock, and the disagreement slowed the project | The date, the day it closed, and what you did to close it |
| "In the end they saw I was right" | A vindication story, which gets probed hard | What you gave up, and what they'd still say |
| A Staff answer: "my team and our PM disagreed" | Senior | Several teams, their incentives, what each gave up |
| "The exec didn't understand the technical risk" | Contempt plus the wrong framing; below Principal | The money, her best reason, what committing looked like |
Step 4: Rehearse the follow-ups
Prepare these five, and say each answer out loud once:
- "What did you do the morning after it closed against you?" The most dangerous one. Name what you did that day. If you won, prepare "What if it had gone the other way?"
- "How long did it take to close, and was that too long?" Say the number of days, and what you'd do to shorten it.
- "What did the other person say about you afterwards?" The honest version.
- "How did your manager see it?" What they said at the time, including any concern.
- "What did you give up?" Something real, and who agreed to it.
Common mistakes
- Telling a relationship story. Pick one where the decision was the hard part.
- Quiet non-commitment. "I went along, but I kept my doubts in the docs" sounds like building a file. Say what you stopped doing.
Key takeaways
- It tests whether disagreement speeds up or slows down the decision, and whether you commit when it closes against you.
- The level shows in who you had to close it with, from a teammate to an executive.
Test yourself on this question's key ideas: