</> MAANG.io

Foundations

The behavioral questions new grads and SDE I candidates are asked, from tell me about yourself to your first mistakes and disagreements. Every lesson also shows how the expected answer grows from SDE to Principal.

0 of 36 solved 0%

Tell Me About a Time You Had a Conflict With a Coworker

The question: "Tell me about a time when you had a conflict with a coworker. How did you handle the situation and what was the outcome?"

Also asked as: "Tell me about someone at work you found hard to deal with. What did you do?" · "When has a colleague been frustrated with you, and why?" · "Tell me about a conflict with a peer leader in another org. Where do things stand with them today?"

Who's asking: an Engineering Director or Senior Engineering Manager, in a behavioral round.

Executive Summary: The director is deciding how you treat people when it's hard, and whether you'll bring drag or trust to their team. The strongest signal is how the other person comes across in your telling: their best argument, your own part, and an honest account of how things stand now. What changes with level is how far the friction reaches: one pull request, a project you own, the goals of competing teams, then two orgs and their money.

What the director is really asking

Underneath the story, the director is answering three questions of their own:

  • "How do they treat someone they're annoyed with?" Contempt for a colleague, even subtle, is a no-hire signal at every level. Directors hear it in small words: "just", "obviously", "to be fair, he's smart, but".
  • "Can they name their own part?" People who can't see their share are expensive to manage: every conflict reaches their manager as someone else's fault.
  • "What happens afterwards?" Conflict is normal. The director wants to know whether you could still work together, and whether you'd tell them the truth if you couldn't.

This question is about a person. A technical debate nobody was upset about answers a different question.

It maps most closely to Amazon's "Earn Trust", and more loosely to "Have Backbone; Disagree and Commit"; treat both as approximate. Three ladder dimensions do most of the work: conflict (who the friction is with, and what you had to broker), how you treat people, and escalation hygiene (who you spoke to first).

Level fit at a glance

SDE Senior SDE Staff Principal
Asked at this level? Core Core Common Common
Story scope they expect Friction with a teammate over your feature or code Friction with a peer, on your team or the next one, hurting a project you own Friction with a lead whose team's goals compete with another team's Friction with a peer leader in another org, with a business cost
The bar in one line Own your part, talk to them directly, and work well together afterwards Go to the peer first, settle it with data, and say honestly where it stands Find the incentive by asking, align the managers, and keep the person's trust Say the hard thing to a peer leader early, price it, and keep working through what stays open
Instant down-level "She was nitpicking", or "I told my lead and he fixed it" "I escalated, and his manager sorted it" Two people on one team, or "she was just protecting her metric" A personal clash with no business cost, or "the VP decided it my way"

It's Core at SDE and Senior. At Staff and Principal, directors lean on influence and dissent questions, but still ask the plain version to hear how you talk about a colleague.

Quick test: could the other person sit beside you while you told it? If you'd change a word with them in the room, change it now.

SDE: own your part, talk to them directly, work well together after

What meets the bar

  • There's a real person and real friction, and you name what you did to make it worse.
  • You went to the person yourself, and you knew when to involve your lead.
  • You can give their side at its strongest, and still keep any fair point of your own.
  • The work shipped with a result you can point to, and you can show what working together looked like afterwards.

Model answer

"I'm a backend engineer on the transcoding team at a podcast-hosting platform, about two years in. I was changing how we retry failed audio transcodes, a ticket we'd sized at two days. By day nine I was on a third round of review with a teammate who'd been on the team four years, and she'd left another long set of comments.

I replied in the pull request, where the whole team could read it: 'If you'd read the ticket, you'd know why this is scoped the way it is. We can't gold-plate every retry.' She stopped commenting. My lead asked if I wanted him to step in. I asked for a day, because the comment was mine.

First I went back to the same thread and wrote that my last comment was out of line and her questions were fair. I wanted the apology where people had read the jab. Then I asked her for fifteen minutes, and said I wanted to understand what worried her, and I wasn't there to defend the code.

She told me something I hadn't known. Before I joined, a retry had published a cut-off episode to about three thousand subscribers, and she'd spent that night on the cleanup. My retry path could do the same thing. I also told her the part that had got to me: I couldn't tell which of her comments were blocking. She said she'd been writing them between on-call pages, and she started marking nits as nits.

I added a check that compares the output's duration to the source before publishing. It merged four days later, so a two-day ticket took thirteen. In its first two months the check stopped 41 truncated files from reaching feeds. My next four pull requests with her took one or two rounds, and that spring she asked me to pair with her on chapter markers.

My habit now is simple. After two back-and-forths on one comment, I stop typing and ask for ten minutes."

Why this lands at SDE

  • "I wanted the apology where people had read the jab": the level beat. The wrong was public, so the repair was too, and the scope is one thread on one pull request.
  • "She'd spent that night on the cleanup": her caution has a reason, and there's no villain. The candidate's fair point is said without blame.
  • "My next four pull requests with her took one or two rounds": the relationship has evidence, beside a result for the code and a stated cost.

Follow-ups to expect

  • "When would you have let your lead step in?": tests knowing when to ask for help → if she still wouldn't talk, or the work was blocked past a date.
  • "Were all of her comments fair?": tests whether the apology was real or just folding → some were nits; you raised that with her in person, not in the apology.
  • "What did she say about you afterwards?": tests whether the repair held → what she actually did, such as asking you to pair, beats what you hope she thinks.

Red flags

  • ❌ "She was nitpicking my code." Directors hear contempt, and wonder who you'd say it about next.
  • ❌ Apologising in private for something you said in public. The team only saw half of it.

Senior SDE: go to the peer first, settle it with data, say where it stands

What meets the bar

  • The friction is with a peer, and it's hurting a project you own, reaching one or two adjacent teams at most.
  • You talk to the peer before their manager. If you didn't, you own that as your mistake, with no process language around it.
  • You settle it with data, even when it goes against you, and say what you gave up.
  • You can say honestly where the relationship stands. Unhealed is fine if it's specific.

Model answer

"Our monthly roster has to reach about 4,000 flight crew by a fixed date. I led the five-person team that validates and publishes it, at a company that makes crew-rostering software for airlines. The optimizer team next door generates the pairings, and a senior engineer there kept shipping changes inside our last 72 hours. Each release forced a 30-hour re-validation, and twice in a row my team worked the weekend.

After the second weekend I went to his manager and asked her to get him to stop shipping inside the window. I hadn't said a word to him. I told myself it was a process problem, but I was tired and avoiding an awkward talk. She raised it in their one-on-one. He came to my desk that afternoon and asked, quietly, why he'd heard it from her. I didn't have a good answer. I said I should have come to him first, and I was sorry.

Then I asked what I'd skipped. Those changes were mostly fixes for rest-rule violations that crew schedulers only found once they saw draft pairings. Holding one for a month meant schedulers hand-repairing around 200 pairings. His position was that crew legality outranks my team's weekend, and I had no answer to that either.

We went through a year of late changes together, and 17 of 23 were legality fixes. The other six could have waited, and he agreed to hold them. For the rest, the cheap thing to change was on my side. I took it to our PM, and we gave up five weeks of the roster-swap feature to make validation incremental. A mid-level engineer on my team built it, the biggest change she'd led, and he started posting pending fixes 48 hours ahead.

Re-validation dropped from 30 hours to about 3. Over the next eight cycles we published on time seven times, against five of the previous eight, and my team worked one weekend.

He and I work fine now. He reviews our contract changes within a day and tells me straight when he disagrees. But he still copies his manager on anything where we might not agree. I taught him to do that."

Why this lands at Senior SDE

  • "I hadn't said a word to him": the wrong sequence, owned as a decision. Then "17 of 23" goes against the candidate, who says so and still holds the point on the other six.
  • "We gave up five weeks of the roster-swap feature": the level beat. A tech lead spends their own team's roadmap, agreed with the PM, to remove the friction. One adjacent team and no managers realigned, so no Staff moves.
  • "He still copies his manager": an ending directors believe more than a happy one.

Follow-ups to expect

  • "Why did you go to his manager first?": tests honesty about your motives → the personal reason, and what you do now before you escalate.
  • "What did his manager think of you after that?": a people probe → what she said, and that you told her you'd skipped him.
  • "What have you done to earn it back?": tests whether you'd oversell → the specific things you've done, and the admission that it isn't finished.

Red flags

  • ❌ "His team was careless about our window." Your weekend is real, but that ignores what he was protecting.
  • ❌ "And now we're great friends." A tidy ending gets probed, and it often doesn't survive.

Quick check before moving on:

Staff: find the incentive by asking, and keep the person

What meets the bar

  • The friction is with one person, but its cause sits between teams, in competing goals and a gap nobody owned.
  • You asked what they answer for, instead of guessing at their motives.
  • You aligned the managers without authority, and you credit the teams for the building.
  • You name a cost you accepted, and you'll defend it.

Model answer

"A better fraud model had been ready for four months at the regional insurer where I was staff engineer for claims systems, and the engineering lead for claims processing had blocked its rollout twice. I sat between the fraud-model team and hers and managed neither. By the second block I'd told both managers she was protecting her cycle-time number.

I was wrong about that. I built a daily cap on flags that held average cycle time flat, and brought it to her as the fix. She said no and stopped coming to the working group.

I asked her for an hour, and started by saying I'd described her motives to her manager without ever asking her. Then I asked what she answered for. She pulled up the prompt-payment report. Flagged claims were 3 percent of volume and almost half of our late-payment penalties and state complaints. Her argument was that a flagged claim is someone with a flooded kitchen waiting on money, and nobody on the fraud side ever sees that person. An average hides that, and my cap did nothing for it.

A flagged claim sat in fraud's queue while claims' legal clock ran, and neither team's goal counted those days. She and I wrote that up together. Then I asked both managers to replace their separate goals for the year with one number: dollars recovered, net of investigation cost and late-payment penalties. The fraud manager pushed back for three meetings, and fairly, since flag volume had funded his team. I showed him net recoveries could fund it too, and he came round. Investigations and legal agreed to a twenty-day clock, after which an undecided claim pays automatically. Fraud and claims built it over seven months.

Net recoveries rose about $2.7 million a year, and penalties on flagged claims fell from roughly $900,000 to $340,000. She and I review each other's release plans now, and she's the one who proposed that.

The clock has a price, and it's inside that net figure. We now pay about $600,000 a year in claims that are probably fraud, because time ran out on them. I argued for that number, and I answer for it whenever the fraud team's results get reviewed."

Why this lands at Staff

  • "I'd described her motives to her manager without ever asking her": the misread incentive, owned as something done to a person.
  • "A flagged claim is someone with a flooded kitchen waiting on money": her strongest argument, in human terms.
  • "Replace their separate goals for the year with one number": the level beat. A Senior solves this inside their project; a Staff engineer changes what two managers measure, with authority over neither, and owns the $600,000 that comes with it.

Follow-ups to expect

  • "What did she say about you afterwards?": tests whether the repair held → that she proposed the release-plan reviews, and what she'd still hold against you.
  • "How did you get the other managers to agree?": tests influence → what the fraud manager stood to lose, and what moved him.
  • "How do you defend knowingly paying $600,000 of fraud?": tests judgment → it roughly pays for itself in penalties; the gain is claimants paid on time; who agreed; what would reopen it.

Red flags

  • ❌ "Two engineers on my team didn't get along, and I sorted it out." Senior scope, told by a Staff candidate.
  • ❌ "She was just protecting her metric." A guess about a person, presented as analysis.
  • ❌ "I got her manager to overrule her." Conflict resolved by authority, and the person is lost.

Principal: say it early to a peer leader, and keep working through what stays open

What meets the bar

  • The other person is a peer leader in another org, and the conflict has a cost in business terms.
  • You name your own part at its true price, and you recommend against your own org's interest when the numbers call for it.
  • Something outlives the moment: a mechanism other leaders use.
  • You can say what's still open between you, and how you work around it without stalling an org.

Model answer

"Two orgs at a mobile carrier with 30 million subscribers were building the same system, and for nine months I knew it and didn't say so to the one person I should have. I was a principal engineer in network. My counterpart in billing and I had worked together for eight years, and he'd stood by me through a bad outage year.

Our 5G plan needed real-time charging. My org was putting usage pricing in the network core, in the call path. His was putting it in the billing platform, the revenue record auditors sign. I only raised it in my org's reviews, softly, because I didn't want to take on a friend, and I told myself annual planning would force it. It did. The CFO found two funding lines for one function and asked our VPs why, in a full room. By then about $8 million of code was scrap, and the launch had slipped two quarters.

The next morning I told him I'd known since spring. He was angrier about that than about the money, with every right. His argument: a pricing error in the core is a misstated revenue line, and network engineers don't carry that. Mine: a billing outage in the path stops data for 30 million people.

We wrote one memo to both VPs and the CFO, with our disagreement priced on each side. It recommended one pricing engine in billing, with a small network quota layer. So I was telling my own VP to cancel a program he'd announced. That took a month and two finance reviews.

My nine months cost most of about $20 million: $8 million written off and roughly $12 million of revenue lost to the slip. One engine saves about $17 million a year. I proposed that every org's plan list the systems it assumes another org will build, and finance now cross-checks them.

He and I still disagree on one thing. When billing is down, the quota layer keeps sessions running on usage we can't bill. He thinks that gives away revenue. I think a network that stops is worse. So each year we each write a page, finance sets the limit, and we both build to it. So far it's gone his way once and mine once."

Why this lands at Principal

  • "I didn't want to take on a friend": the correction is a personal failure with an org-level bill, told straight.
  • "A pricing error in the core is a misstated revenue line": his strongest argument, in the terms a CFO cares about.
  • "I was telling my own VP to cancel a program he'd announced": the level beat. A recommendation against the candidate's own org, at exec level, with the $20 million priced as the cost of their silence.
  • "He and I still disagree on one thing": a standing way to disagree that stalls no one. The plan cross-check is the mechanism other leaders use.

Follow-ups to expect

  • "Why did you wait nine months?": tests honesty → the personal reason, and what you do now when two plans look incompatible.
  • "How did you bring your VP along, and what did he say about you?": tests exec influence → what he objected to, and what it did to your standing.
  • "Isn't an unresolved disagreement between two principals a leadership failure?": tests judgment → why this one can stay open, who decides each year, and what would make you escalate it.

Red flags

  • ❌ "The VP decided it my way." Authority ended it, and it shows nothing of your influence.
  • ❌ "Billing just doesn't understand networks." Contempt for another org costs more the higher you go.
  • ❌ "We're fully aligned now." Directors know two senior people rarely are.

Quick check before moving on:

One story, four altitudes

SDE Senior SDE Staff Principal
Situation A sharp public reply on a third review round A peer's late fixes costing my team weekends A fraud model blocked twice by the claims lead Network and billing orgs both building real-time charging
Your role Engineer on the transcoding team Tech lead of the five-person roster team Staff engineer between two teams, managing neither Principal in network, facing an eight-year colleague
The key move Apologised in the same thread, then asked what worried her Apologised for going around him; spent five weeks fixing our side Asked what she answered for; put both managers on one net number One priced memo; told own VP to cancel his program
Result 41 truncated files stopped; next four PRs in one or two rounds Re-validation 30 hours to about 3; on time 7 of 8 cycles, up from 5 Net recoveries up $2.7M a year; flagged-claim penalties $900K to $340K One engine saves $17M a year; my silence cost most of $20M
Reflection After two back-and-forths, ask for ten minutes He still copies his manager; I taught him that I argued for $600K a year of likely fraud Still disagree; each writes a page, finance sets the limit

Draft your answer

Step 1: Pick the right story

Run each candidate story through the test for your target level:

  • SDE: Was it friction with one teammate over your work, did you go to them yourself, and can you show how you worked together after?
  • Senior SDE: Did it hurt a project you own, did data settle it, can you say what you gave up, and can you say where the relationship stands?
  • Staff: Did at least two teams change what they did, did their managers agree to change goals, priorities or ownership, and can you state the other person's incentive as they would?
  • Principal: Is the other person a leader in another org, is there a business cost you can price, did the orgs change how they work, and can you say what's still open?

Then check that it passes all of these:

  • It's from the last two or three years, and you were central to it.
  • There was real friction between two people, and you can name your own part without adding "but".
  • They'd recognise your version.
  • There's a number for the work, and something observable about how you work together now.

If you're a new grad. A team project, a research lab or a part-time job counts if something real hung on it, like a grade, a demo or a submission, and you went to the person yourself, not to a professor or boss. The director accepts a smaller stake, but still wants your own part, a repair you led and how you worked together afterwards: "Two weeks before our capstone demo, I rewrote my teammate's parser without asking him."

Aiming a level above your current title? A Senior aiming at Staff can tell the cross-team version only if the managers really agreed to change goals, priorities or ownership. If no story clears the level, tell your best story at your real altitude, and use the reflection to show you see the next one: "If it came up now, I'd ask both managers what each is measured on before I said anything about the design." An honest Senior story with that awareness beats an inflated one that falls apart at "What did her manager say?".

Step 2: Fill the skeleton

SITUATION       (15s) — Your scope, and who the other person was (role
                        only): ______
FRICTION        (15s) — What happened between you, and why it mattered
                        to the work: ______
YOUR PART       (15s) — What you did that made it worse, with no "but": ______
THEIR SIDE      (15s) — Their strongest argument, in words they'd accept: ______
REPAIR          (35s) — 2–3 moves, each "I …, because …". Who you spoke to,
                        and in what order: ______
LEVEL BEAT (inside REPAIR) — the line only your level could say.
                        SDE: how you repaired it where the harm was done.
                        Senior: what you gave up from your own team's plan.
                        Staff: what the managers agreed to change, and the
                        cost you accepted. Principal: the call you took to
                        execs, even against your own org, and the mechanism
                        other leaders now use.
RESULT          (15s) — One number for the work, its cost, and one thing you
                        can observe about how you work together now: ______
WHERE IT STANDS (10s) — An honest state of it: a plain new habit, what still
                        isn't healed, or what's still open and how you work
                        around it: ______

Aim for about two minutes, two and a half at most, with a 30-second cut ready (the friction, your part, their side, one number, where it stands).

Step 3: Calibrate it

If your answer says… It reads as… Fix
"She was difficult", or "he was just protecting his metric" Contempt, a no-hire risk at any level Their best argument, the way they'd say it
"I told my manager and it got sorted" The action was escalation; below the bar at any level Who you spoke to first, and how your manager came in
"And now we're great friends" Too clean, and it will get probed The specific, honest state of things
A Staff story about two people on one team Senior The teams, the incentives, and what their managers agreed to change
A personal clash with no business cost, told at Principal Staff or below The price, the org change, and what's still open

Step 4: Rehearse the follow-ups

Prepare these five, and say each answer out loud once:

  1. "If I called them today, what would they say about working with you?" The most dangerous one. Say what they'd praise, what they'd still criticise, and how you know.
  2. "What was your part in it?" One specific thing you did, stated as a decision.
  3. "How did your manager see it?" What they said at the time, including any concern, and what you did about it.
  4. "What did you give up?" The feature, the point you conceded, or the cost you accepted.
  5. "Would you work with them again?" Yes or no, with what you'd do differently from day one.

Common mistakes

  • Telling a debate in place of a conflict. Pick a story where the relationship was at risk.
  • Politely hiding a villain. "He's brilliant, but" still makes someone the problem. Make their argument strong enough that a listener could side with them.
  • Your part as a humble-brag. "I cared too much about quality" isn't a part. Name the thing you did.
  • Stopping at the relationship. "We get on well now" isn't a result. Add a number for the work, and what the repair cost.

Key takeaways

  • It tests how you treat people when it's hard, and the director hears it in how the other person comes across.
  • The level shows in how far the friction reaches.
  • You don't need a healed ending, only a true one.

Test yourself on this question's key ideas:

Sign in to MAANG.io

Use your Google or Microsoft account — no password to remember.

Continue with Google Continue with Microsoft

Please accept the terms above to continue.