Tell Me About a Time Your Work Was Criticized
The question: "Tell me about a time when your work was criticized. How did you handle it and what did you learn?"
Also asked as: "Tell me about a time your design was shot down." · "When did someone last tell you your work wasn't good enough?" · "How do you handle being criticized in front of other people?" · "Tell me about a time an executive or peer pushed back hard on a direction you'd set."
Who's asking: an Engineering Director or Senior Engineering Manager, in a behavioral round.
Executive Summary: The director wants the moment the criticism landed, usually in front of other people, and what you did in the next ten seconds, the next day and the next quarter. The best answers separate what was right from what wasn't, act on both, and carry the cost.
What the director is really asking
The director has given hard feedback to hundreds of engineers and watched three things happen: they argue, they fold, or they sort it.
- "What did they do in the room?" What they said, and who was listening. If you argued, the director would rather hear that and the next morning's recovery than a story where you were calm throughout.
- "Could they tell the right part from the wrong part?" Criticism is almost never entirely fair. Agreeing with all of it reads as folding; disputing all of it reads as defensive. The director listens for one sentence on each.
- "What did it cost, and what changed?" A number for the fix, a number for its cost, and something you now do differently that another person could confirm.
The feedback arrived; you didn't go looking for it (that's a different lesson). From Senior up there's a fourth: "Can this person be wrong in public at the scope I'm hiring for?"
It maps most closely to Amazon's "Earn Trust" (vocally self-critical) and, because part of the criticism is usually wrong, to "Have Backbone; Disagree and Commit". Both approximate. The ladder dimensions at work are people, conflict (who the critic is) and the failure story (the cost named and carried).
Level fit at a glance
| SDE | Senior SDE | Staff | Principal | |
|---|---|---|---|---|
| Asked at this level? | Core | Common | Common | Common |
| Story scope they expect | Your feature, criticized by one person in front of your team | A system decision you own, criticized by a partner director with your manager present | A cross-team proposal, criticized in public by leads you don't manage | A direction you set, criticized by an executive in front of the exec team |
| The bar in one line | Hear it, look at the examples, fix what was right, say what wasn't | Don't argue the number; change the right-sized thing; own the cost; change a team practice | Hold the problem, concede your move, get other managers to change something they own | Don't win it at the table; work it through with the critic; change your own plan; leave a mechanism |
| Instant down-level | "I agreed and fixed everything they said" | "I showed the director her number was wrong" | "My manager pushed back on my design" | "A VP challenged my design and I presented the data" |
Core at SDE because coachability is the primary signal with little track record; from Senior up, the critic has power over you.
SDE: hear it before you defend it
What meets the bar
- You can say what they said, close to their words, and who else was in the room.
- You're honest about the room. If you argued, say so, and what you did next day. A recovery beats serenity.
- You went and looked at the examples before deciding what was right, fixed that at a stated cost, and said plainly what was wrong.
Model answer
"I work on the publishing pipeline at a daily crossword app, and about two years in I built a preflight validator that checks a constructor's puzzle file before the editors schedule it. I'd tested it against three months of standard grids and it passed all of them.
In our weekly editorial sync, the head puzzle editor said it had rejected 11 of the 14 puzzles they'd tried on Tuesday, that the messages were unreadable, and that it had lost one. The editors had gone back to checking by hand. Two editors and my whole team were in the room, and I defended it. I said the rules came straight from her style guide, and that it hadn't lost anything, the file was still in the queue. I was right about the file. My lead caught me afterwards and said, 'You were right about the file, and it didn't matter.'
I booked fifteen minutes with her the next morning and said, 'I argued with you yesterday instead of listening. Can you show me the eleven?' Seven were themed puzzles with deliberate asymmetry, which editors allow and I'd treated as an error. Three were a bug in my clue-length check, which counted an em-dash as two characters. One was genuinely broken, and she said that catch was the only reason they'd kept trying. So ten of the eleven were mine, and each message read something like 'symmetry constraint failed at (3,7)'.
I made asymmetry a warning the editor can dismiss, fixed the count, and rewrote every message to name the clue or cell and say what to do. She read every message before I shipped. That took about a week and a half, and the preview-image work I'd planned slipped two weeks.
Over the next month about sixty puzzles went through it. It flagged six, and the editors said five were real. They stopped checking by hand, which she put at twenty minutes a puzzle. Now when something I built is criticized, I ask to see the examples before I say a word about whether it's right."
Why this lands at SDE
- "I booked fifteen minutes with her the next morning and said, 'I argued with you yesterday instead of listening. Can you show me the eleven?'": the line only an SDE would say, because of its scope: one critic, one feature they built, and the recovery itself is the whole story. Nothing about how a team works had to change.
- "I was right about the file. My lead caught me afterwards": they name the wrong part and still see that defending it was the mistake.
- "ten of the eleven were mine": the examples decided, not the argument.
Follow-ups to expect
- "You were right about the file. Why was that the wrong thing to say?": the most dangerous one → it tests whether you see that being right on the small point cost you the room.
- "What did your lead say afterwards?": tests honesty about the defensive moment → quote it, and what you did with it.
- "What did you learn that you still use?": the SDE probe → the examples-first habit, and when you last used it.
Red flags
- ❌ "I thanked them for the feedback and fixed everything they raised." No judgment. Ten of eleven were yours; one wasn't.
- ❌ "I explained why the validator was correct and they understood." Nothing was learned.
Senior SDE: change the right-sized thing
What meets the bar
- The work criticized is a system decision you own, and the critic has standing, often with your manager present.
- You didn't argue the number in the room; you got the real one.
- You handled the right part and the wrong part differently, and own the cost of your response, including if you changed too much.
- You changed a practice on your team.
Model answer
"At a fuel-card company for trucking fleets, I led a team of five on fraud detection. I'd designed a risk-scoring model that auto-declines a fuel purchase above a threshold, and fraud losses fell by about $210,000 a quarter.
At the quarterly review, with my manager and the head of risk in the room, the customer-success director said, 'Your system declined 1,900 legitimate fills last month. Drivers are standing at a pump at two in the morning with a card that doesn't work, and your team isn't taking the call.' Her number was wrong. 1,900 was every decline, including real fraud and rules that predated me. But I'd never looked at false declines, and I'd never spoken to the call desk. I didn't argue in the room. I said I'd bring the breakdown within a week, and I did: my scoring was adding about 640 false declines a month. I corrected the 1,900 once, in that follow-up, and left it there.
Then I overcorrected. In one week I raised the threshold and moved our two strongest rules, tank capacity and distance from the truck's telematics, from decline to flag. False declines fell about 80 percent, and fraud came straight back: about $95,000 in three weeks that the old rules would have caught, most of it tank capacity. I told the head of risk and my manager I'd overcorrected, with the number, and put tank capacity back to a hard decline. For the distance rule I did two overnight shifts with the call desk, and with their lead replaced the decline with a text to the driver and a 90-second confirm.
I set up two numbers my team reviews side by side every week, fraud caught and false declines, and no scoring change ships without an estimate for both. The customer-success director gets the same dashboard. Two quarters on, false declines were about 180 a month and fraud losses within about 10 percent of our best month. It cost $95,000 and three weeks of churn. That first week I was fixing how I felt, not the system. Now I change the smallest thing that answers the criticism, measure, then decide on the next."
Why this lands at Senior SDE
- "I set up two numbers my team reviews side by side every week, fraud caught and false declines, and no scoring change ships without an estimate for both": the line only a Senior would say. A practice on their own team; nobody else's roadmap changed.
- "Her number was wrong … But I'd never looked at false declines": wrong part, right part, handled differently.
- "That first week I was fixing how I felt, not the system": they own the overcorrection and price it.
Follow-ups to expect
- "You lost about $95,000 fixing a complaint. What did the head of risk say?": the hardest one here → don't defend it. Say what he said, that you told him first, and what the weekly review now catches.
- "What did you trade off?": tests judgment → fraud loss against stranded drivers, and what you count now.
- "Who did you grow on this?": the Senior probe → who runs the weekly review now, and what they've changed.
Red flags
- ❌ "I showed the director her number was off by a factor of three." You won the number and lost the room.
- ❌ "Customer success should have come to us first." You've made the critic the villain.
Quick check before moving on:
Staff: criticized in public by people you don't manage
What meets the bar
- The work is a proposal spanning several teams for a problem nobody owned; the critics are leads you don't manage, in public.
- You held what was right about the problem, conceded what was wrong about your move, and found out what the criticism was actually about.
- Their managers changed something they own, each for a reason of their own, and you can say what you stopped doing.
Model answer
"At a national lottery operator, five teams touch a draw: terminals, the draw, prize claims, player accounts and regulatory reporting. Reconciling each draw against sales and claims took reporting three days and had produced two regulator queries in a year. Nobody owned that, and I'd sat in both postmortems, so I wrote an RFC: one event schema across the five teams, with a two-quarter adoption date for each, and posted it company-wide.
Within a day, three team leads were taking it apart in a forty-comment thread. The terminals lead said it doubled the payload on a slow link to 9,000 terminals. The prize-claims lead said my mandatory ticket id didn't exist for about 30 percent of high-tier claims, which arrive on paper first. The player-accounts lead wrote, 'You've put us in Q3, mid-migration. Who decided this?'
I treated it as a technical review and spent two weeks on version two. The thread went quiet and nobody committed a date. I only understood when the player-accounts manager asked my director whether 'this schema thing' was funded, and the terminals lead told me he'd have found the payload issue with me in a room. He'd commented because I'd put a date on his team in public.
I withdrew the timeline, kept the schema, and went to the three managers one at a time to ask what adopting would cost them and when it would be cheap. Terminals wanted the compact encoding and a half-year. Prize claims wanted paper-first claims as a two-event sequence, which I brokered with reporting. Player accounts wanted to go after their migration, so reporting carried a shim for two quarters, about three weeks of their time. I handed my own team's design reviews to its senior to make room.
The other four teams adopted over three quarters, one later than my timeline; my own team's draw system went first. Reconciliation went from three days to about four hours, and the next year had no regulator queries. It cost the shim, a quarter, and two weeks answering the wrong objection first. The comments were about the schema; the criticism was about the dates. Now when a group takes work apart in public, I find out what each needs to say yes before I touch the document."
Why this lands at Staff
- "I withdrew the timeline, kept the schema, and went to the three managers one at a time to ask what adopting would cost them and when it would be cheap": the line only a Staff engineer would say. Managers who don't report to them change dates and carry work; the candidate gave up the part they got wrong.
- "The comments were about the schema; the criticism was about the dates": the real objection, and an admission they missed it for two weeks.
- "about three weeks of their time": the cost lands on a named team.
Follow-ups to expect
- "Why was it yours to publish a timeline for three teams you don't manage?": the hardest one here → it wasn't, and that was the criticism. The problem was yours to frame; the dates were theirs.
- "How did you get the other managers to agree?": the Staff probe → what each wanted and paid.
- "What did you stop doing to make room for this?": the Staff probe → your own team's design reviews, and what the senior who took them changed.
Red flags
- ❌ "I escalated to the director, who set the adoption dates." The teams complied; nobody agreed.
- ❌ "The leads were defending their turf." Contempt for the critics, who were right about the dates.
Principal: work it through with the critic before the table
What meets the bar
- The work criticized is a direction you set; the critic is an executive, in front of the exec team, with funding at stake.
- You didn't try to win it at the table, or you tried once and say what it cost.
- You worked it through with the critic, on their ground, and changed your own plan where they had a point, at a cost to you.
- The org kept a mechanism.
Model answer
"I'm principal engineer at a national blood-donation service. I'd set a two-year direction: one shared donor-and-donation platform replacing three systems, about $7 million. Appointment, testing and inventory data only met in a nightly batch, so a rare-type shortage took about 36 hours to become targeted outreach.
In the exec review, in front of the CEO, CFO and medical director, the COO, who runs collections, said: 'A two-year IT programme that gives me nothing until month 18. My fourteen regional centres run drives on a system that falls over weekly, and your plan tells them to wait. I'm not funding my half.' Part of that was wrong: donor outreach went live at month eight. The rest was right. I'd sequenced by data flow, collections was the last hop, and I'd never spent a day at a drive.
Then I made my mistake. I built a rebuttal around the month-eight milestone and took it to the next review. She brought her own numbers: about sixty drives cancelled the previous year for system outages, roughly 4,000 units. The CEO said, 'Come back with one plan.' Funding slipped a quarter, and the medical director, who'd backed me, went quiet.
I asked the COO for a day, spent two at her drives, then three weeks re-sequencing with her and her head of collections. I pulled her collections scheduler forward to month eight and pushed my lab integration back six months, and I took that cost to the medical director myself. We presented one plan, and she funded her half. And I changed how direction is reviewed: any platform plan goes to the affected exec owner two weeks before exec review and comes back with their sign-off or written objection. The CFO's planning office owns that now.
The scheduler went live in month eight, and drive cancellations for system reasons fell from about sixty a year to nine. Shortage-to-outreach only reached 20 hours at month 20, six months later than my plan, and one rare-type shortage in that gap ran 40 hours, which the medical director has not let me forget. The deck was right, and it cost a quarter. Now I work a criticism through with the critic before I answer it in the room where they made it."
Why this lands at Principal
- "I pulled her collections scheduler forward to month eight and pushed my lab integration back six months, and I took that cost to the medical director myself": the line only a Principal would say. A funded direction changed against their own preference, and they carried the cost to the exec who paid it.
- "Then I made my mistake. I built a rebuttal": the sequencing error is named, priced at a quarter, and not blamed on the COO.
- "any platform plan goes to the affected exec owner two weeks before exec review": a mechanism owned by someone else, applied to the candidate first.
Follow-ups to expect
- "The CEO sent you both away. Would you have gone to the COO without that?": the hardest one here → be honest, and say the pre-review mechanism exists because you don't trust yourself under pressure.
- "What did the medical director say when you pushed their part back six months?": tests how you carry a cost → the conversation, and the 40-hour shortage.
- "What mechanism outlived your involvement?": the Principal probe → the exec-owner pre-review, who owns it, what it misses.
Red flags
- ❌ "I presented the data and the executive team sided with me." You won a vote and lost a COO.
- ❌ "The COO didn't understand the architecture." Contempt for the critic, who was right about her centres.
Quick check before moving on:
One story, four altitudes
| SDE | Senior SDE | Staff | Principal | |
|---|---|---|---|---|
| Situation | Your validator rejected 11 of 14 puzzles; the head editor said so in front of your team | Your fraud scoring stranded drivers; a partner director said so to your manager | Three leads took your cross-team RFC apart in public | The COO refused to fund your direction in the exec review |
| Your role | Built the feature | Led the team of five; designed the scoring | Staff; no authority over the five teams | Set the direction; accountable for the $7M bet |
| The key move | Argued, then asked to see the eleven the next morning | Didn't argue the number; overcorrected; pulled the capacity rule back | Withdrew the dates, kept the schema, asked three managers what adopting would cost | Took the rebuttal to the table, then two days at her drives and re-sequenced |
| Result | 5 real flags of 6 on 60 puzzles; 20 min a puzzle saved; two-week slip | False declines 640 → 180 a month; about $95k of fraud and three weeks lost | Three days → four hours a draw; no regulator queries; a quarter and a shim | Cancellations 60 → 9 a year; lab integration six months late; a quarter lost |
| Reflection | See the examples before you defend | Change the smallest thing, measure, then decide | The comments were about the schema; the criticism was about the dates | Work it through with the critic before the room they said it in |
Draft your answer
Step 1: Pick the right story
Run each candidate story through the test for your target level:
- SDE: Your own feature, criticized by someone with standing, and you can say what you did in the room and the next day.
- Senior SDE: A system decision you owned, criticized by a partner or director in front of your manager, and your team changed a practice because of it.
- Staff: A proposal across several teams, criticized in public by leads you don't manage, and at least two of their managers changed something they own.
- Principal: A direction you set, criticized by an executive with funding at stake; the direction changed at a cost to you, and the org kept a mechanism.
Then check all of these:
- It's from the last three years, and you can quote the criticism close to their words.
- Part of it was right and part wasn't, and you can say what you did about each.
- You can say what you did in the room, including if you argued.
- There's a number for the fix and one for what it cost.
If you're a new grad. An internship code review, a maintainer rejecting your pull request or a capstone sponsor's verdict all count, if the work was yours and the criticism was specific. A director accepts a mentor or maintainer as the critic and smaller work, but still wants your first reaction, the part you thought wrong and a habit you changed. "A maintainer closed my first pull request with one line: 'This breaks every Windows user.'"
Aiming a level above your current title? Tell your own-level story honestly and show you see further. A Senior aiming at Staff can tell the fraud-scoring story and add: "If that had come from three teams instead of one director, I'd have gone to their managers first, because the fix wasn't mine to schedule." That beats a borrowed cross-team story that collapses at "Why was this yours?".
Step 2: Fill the skeleton
SCOPE + THE WORK (15s) — Your product, your role, the piece of work: ______
THE CRITICISM (20s) — Who said it, who was listening, their words: ______
IN THE ROOM (15s) — What you said in the next ten seconds. If you argued,
say so, and what you did the next day: ______
RIGHT / WRONG (20s) — The part that was right; the part that wasn't: ______
ACTIONS (30s) — 2–3 moves, each "I …, because …", including the one
you got wrong: ______
LEVEL BEAT (inside ACTIONS) — the line only your level could say.
SDE: the examples you looked at, and what you changed.
Senior: the practice your team changed, and what it cost.
Staff: what you withdrew, and what each manager changed.
Principal: what moved in your direction, and the mechanism kept.
RESULT (15s) — One number, and what it cost: ______
LEARNED (5s) — One line on what you do differently now: ______
Aim for about two minutes, with a 30-second cut ready. For the Senior story: "A partner director said my fraud model stranded drivers. Her number was high, but I'd never measured false declines. I overcorrected, lost $95k, pulled it back, and now my team reviews both numbers weekly."
Step 3: Calibrate it
| If your answer says… | It reads as… | Fix |
|---|---|---|
| "I agreed straight away and fixed everything they raised" | No judgment; you folded | The part that was wrong, and what you did about it |
| "They were wrong, and I showed them the data" | Defensive; nothing learned | The part that was right, and what changed in your work |
| "I didn't take it personally", and nothing else | You didn't hear it | What you said in the room, including if you argued |
| "I learned to communicate better" | Nothing changed | What you did differently next time, with a number |
| A Staff answer: "my manager pushed back on my design" | Senior | Several teams' leads, in public; their managers changing something they own |
| A Principal answer: "a VP challenged my design, so I presented the numbers" | Staff | A direction or funding decision moved, at a cost to your plan; a mechanism |
Step 4: Rehearse the follow-ups
Prepare these five, and say each out loud once:
- "What did you say in the room?" The honest version. If you argued, the recovery is your answer.
- "Which part of the criticism was wrong, and what did you do about it?" One correction, delivered once.
- "What did it cost to fix?" The slip, the dollars, the team time. A free fix sounds invented.
- "If I asked the person who criticized you, what would they say now?" Their words, not yours.
- "What do you do differently now?" One habit, and the last time you used it.
Common mistakes
- Choosing criticism that cost you nothing. A nitpick you fixed in an afternoon tells the director nothing. Pick the one with a room, a cost and a part you got wrong.
- Agreeing with all of it. If you can't name the part that was wrong, the director assumes you can't tell.
- Winning the wrong part, or blaming the critic. Correct an error once, with the real number, and move on. "They didn't understand" is contempt, and a no-hire signal.
- Ending on the lesson. "I learned to listen" is nothing. End on what you do differently, in a way another person could confirm.
Key takeaways
- It tests the next ten seconds, then whether you could separate the right part from the wrong part and act on both.
- As level rises, the critic gains power and the room gets bigger, so winning the argument costs more.
- Every response has a cost. The best answers price their own overcorrection, delay or slip.
- "What did you say in the room?" is the follow-up to prepare first.
Test yourself on this question's key ideas: