</> 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 Recovered From a Significant Setback

The question: "Tell me about a time when you faced a significant setback in your work. How did you recover from it and what did you learn?"

Also asked as: "Tell me about a time a project you cared about got cancelled or fell apart. What did you do next?" · "When has something outside your control wrecked your plan? What did you do that first week?" · "Tell me about a time you lost people, a vendor or a dependency partway through a project."

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

Executive Summary: The director is deciding what you do when a plan dies for reasons you didn't choose: whether you regroup and bring people with you, or wait, rush or sulk. What grows with level is who has to believe the new plan: you and your lead, your team, several managers, then the executives who funded the old one.

What the director is really asking

Plans die: a dependency shuts off, a layoff takes half the team, a supplier disappears. None of that says much about you; what you did next does. A director listens for:

  • What died, in one breath. What ended the plan, with no villain and no self-pity.
  • The first week, in order. What you found out, what you stopped and who you told. "We regrouped" hides all of it.
  • What you saved, and what's gone. A dead plan rarely comes back whole, so "we got back on track" gets probed.
  • Who came with you. Who lost something, what they heard from you, and when.
  • What you'd do sooner. Something you'd do on day one next time, not "I learned to be adaptable".

This isn't the mistake question. Here, you didn't cause it. If your story has a wrong turn, it belongs in the recovery: you held on too long, went quiet or grabbed the first way out.

The decision it feeds: "when a plan dies on this person's watch, will people still follow them to the next one?" It maps loosely to Amazon's "Deliver Results", whose wording mentions rising to the occasion despite setbacks; approximate. It turns on ownership, delivering through others and escalation hygiene.

Level fit at a glance

SDE Senior SDE Staff Principal
Asked at this level? Common Common Common Common
Story scope they expect An outside change takes away what your component relied on Your team's project loses people, money or its date A failure you didn't cause stops several teams and reaches customers A funded direction overturned by something the company doesn't control
The bar in one line Let go in time, ask, and ship what still works Re-plan for the team you have, and tell that team straight Get several teams moving, and test the way out on your real situation Give executives a reset they can believe, then leave a mechanism behind
Instant down-level "The API was shut off, so my lead re-planned" "We all pulled together and worked weekends" "My team found a new vendor" "The regulator changed the rules, so we pivoted"

Common, not Core: "your biggest failure" usually takes the slot. At Principal it often arrives as "a strategy overturned from outside", and you'll answer for how the bet priced that risk.

SDE: let go in time, and keep what still works

What meets the bar

  • Something outside your team took away something you built, told without grievance.
  • You stopped trying to save it, and went to your lead early enough for that to help.
  • You kept what still worked, with a number for what got worse.
  • Calls that weren't yours stayed with their owners: the PM on cost, counsel on the terms.
  • Check your altitude: "my lead re-planned and I did my tickets" reads below the bar. If you reset your team's quarter and took the new date to stakeholders, it reads as Senior.

Model answer

"'We will retire our public API on June 30th.' That email reached us on a Tuesday in April, ten weeks out. About two years in at a marketplace for independent knitting-pattern designers, I owned the yarn-substitutes panel: open a pattern, see which other yarns would work. About 12,000 buyers opened it each week, and its yarn data all came through that API.

I spent the next five weeks trying to keep it alive. That first week I wrote to their partnerships address and got 'we'll look into it'. So I built a local cache of all 60,000 yarns we'd pulled, on the theory that we could keep serving them. That took three weeks, and I liked it.

My lead knew about the email and the cache, and had assumed I had a fallback. In week five I sat down to write my status note and couldn't finish the sentence 'if they say no, we…'. I had two answers I didn't control and no plan for either, so I took that to my lead that morning. My lead asked counsel to read the terms, and by Friday we knew: cached data had to be deleted when access ended.

So I priced what was left. A paid catalog quoted $40,000 a year, and the PM said the panel didn't earn that. But designers already name the yarn they used, with its weight and yardage, when listing a pattern, and 78 percent of listings had all three. I rewrote the matching against yarns named across our own listings: same weight class, enough yardage. I hid the panel where the fields were missing, and wrote the note the PM sent designers asking them to fill those in. I shipped at the end of week nine and deleted the cache.

It's a cruder panel. It can't tell alpaca from acrylic. Opens settled around 11,000 a week, clicks on suggestions fell from 14 percent of opens to 9, about 40 percent fewer each week, and a search filter I'd promised slipped a quarter. When a dependency gives me a shutoff date, I read the terms that day and give saving it one week. No written yes by then, and I'm building the replacement."

Why this lands at SDE

  • "I had two answers I didn't control and no plan for either, so I took that to my lead that morning": the line only an SDE would say. They know where their part stops, and they brought it up themselves.
  • "It can't tell alpaca from acrylic": the loss, in the user's terms, with 14 to 9 behind it.
  • "No written yes by then": the lesson is a date someone could check.

Follow-ups to expect

  • "You had ten weeks. Why did it take five to tell your lead you had no plan B?": the most dangerous one → don't defend the cache. Say what you were telling yourself, and what you'd raise in week one now.
  • "How did you decide when to ask for help?": tests when you escalate → the moment the plan rested on answers you didn't control.
  • "Did buyers notice it got worse?": tests honesty about the loss → 14 to 9, what you looked at, and what the PM chose to do about it.

Red flags

  • ❌ "They shut it down with barely any notice." Ten weeks is notice, and the story is now about them.
  • ❌ "I worked nights and saved the feature." Working nights isn't the recovery.

Senior SDE: re-plan for the team you have, and tell them first

What meets the bar

  • Your team's project lost something you couldn't prevent, and you own what happened next.
  • The re-plan is arithmetic: work left against people left, with what's cut named and its owner told.
  • Your team hears it from you, soon and straight.
  • The team's process changed, not only your own habit.
  • Check your altitude: "my manager told me what we were cutting" reads as SDE. If you aligned several managers across teams, it reads as Staff.

Model answer

"By noon on a Thursday in February, three of the six engineers on my team had been laid off. The company rents tents, tables and dance floors through 14 depots, and after a weak holiday season it cut about 15 percent of its staff. I was tech lead on the team replacing the booking system, which could sell the same tent twice: 52 double-bookings the summer before. The date was May 1st, before the season.

We had about 60 engineer-weeks left, and three engineers for 11 weeks is 33. I owned the date, so by Monday I'd gone to the PM and the depot operations lead with what three people could carry. Route planning, 18 weeks, was out. Customer self-service, eight, was out. Operations chose its six busiest depots for May, and the other eight would wait for October, another ten. That left 24.

Then I went back to my team and said the plan hadn't changed. That wasn't true. They'd just watched friends walk out, and I told myself they'd had enough for one week. I asked the PM and my manager to let me tell them, and then I didn't: after a week, telling them meant admitting what I'd said. So for three weeks one engineer kept building route planning I had already cut, and all three could tell something was off. In week three, one of them told me they'd taken a recruiter's call, because a lead that calm after layoffs must know something worse.

I told them that afternoon: what I'd agreed, when I'd agreed it, and what I still didn't know about further cuts. Then I asked them to tear the plan apart. They moved the inventory import to the front, because only one departed engineer had really understood it.

We cut over six depots on April 24th, and double-bookings went from 52 to 27 that summer, nearly all at the other eight, which moved in October. Route planning waited a year, so drivers spent that season on a spreadsheet. All three stayed through the season. Our team agreement now says any change to scope, date or people reaches the whole team within one working day, with what we know, what we don't and when they'll hear next."

Why this lands at Senior SDE

  • "I owned the date, so by Monday I'd gone to the PM and the depot operations lead with what three people could carry": the line only a Senior would say. It's their project and their call, with one adjacent partner and no managers to align.
  • "one engineer kept building route planning I had already cut": what the false line cost, in work and in trust. The motive was kind; the statement was still false.
  • "Our team agreement now says": a team process with a time limit, the Senior ending.

Follow-ups to expect

  • "You told your team the plan hadn't changed, and it had. What do I make of that?": the most dangerous one → don't soften it to "I held back". Say it wasn't true, what it cost, what you did that afternoon, and that you corrected it only when someone was close to leaving.
  • "Your team asks 'are we next?' and you don't know. What do you say?": tests how you treat people → what you know, what you don't, and when they'll hear next.
  • "How did your manager see the re-plan?": tests whether you escalated cleanly → what your manager asked, and whether you told them you'd sat on it.

Red flags

  • ❌ "I shielded the team so they could focus." That's the story's wrong turn, told as a virtue.
  • ❌ "We worked weekends and hit the date." With half a team, a date that holds without cuts isn't believable.

Quick check before moving on:

Staff: move several teams, and test the way out first

What meets the bar

  • A failure crossing several teams, with customers feeling it, and a reason it landed on you.
  • "I" is the plan and the alignment; the teams built it.
  • The way out was tested on your real situation, or you reverse it yourself when the evidence comes in.
  • Friction: who said no first, and what changed their mind.
  • Check your altitude: your team finding a new vendor reads as a big Senior project. A funded direction priced in revenue and margin, in front of the board, reads as Principal.

Model answer

"On a sunny Monday in June, 60,000 homes' apps showed their panels had produced nothing. The panels were fine. I'm a staff engineer at a home-solar monitoring company, and those homes' gateway supplier had gone insolvent over the weekend, taking its cloud with it. Procurement held the contract, firmware the device, pipeline the data and mobile the app, each waiting on another's answer, so I told my director I'd write the recovery plan. The teams did the building.

Support took 9,000 calls that week. I asked mobile for a banner by Tuesday: panels fine, data delayed. I asked counsel whether escrow covered the firmware. By Thursday I'd recommended the first replacement that could ship: a new gateway mailed to every home at $70, with 10 percent needing a $200 technician, on that supplier's figures. About $5.4 million. My director and finance signed the full order. The first 2,000 shipped; nobody called it a test.

By week five, only 52 percent were online. I went through the failures with the installer operations lead: about 40 percent of our gateways sat in garages and attics on cellular, and the new box was Wi-Fi only. That figure had been in our own fleet data. At 48 percent failing, the plan cost about $10 million, so I told my director I'd recommended it too fast and asked to stop the order.

In week seven, counsel got the signing keys from escrow and the hostname the gateways called home to, so we could reflash the boxes already on walls. The firmware manager said no first: a bad flash across 60,000 homes was worse than the outage. That manager agreed once we staged it, 500 homes, then 5,000, with the stop theirs to call.

By week 18, 88 percent had reflashed. The other 7,200 got a technician and a new box: with the pilot, about $2.1 million. Most homes had no data for 10 to 18 weeks, the last few thousand until about week 24. It took 70 engineer-weeks, and a battery launch slipped a quarter. I don't back a replacement now without a success rate from 500 of our own installs, sampled by how they're really wired, however fast the calls are coming in."

Why this lands at Staff

  • "each waiting on another's answer, so I told my director I'd write the recovery plan": the line only a Staff engineer would say. No one assigned it, and it answers "why was this yours?" before it's asked.
  • "That figure had been in our own fleet data": the first way out failed on something checkable. They say so, and ask to stop it themselves.
  • "It took 70 engineer-weeks, and a battery launch slipped a quarter": the recovery's own cost.

Follow-ups to expect

  • "Testing on 500 homes takes weeks, with 9,000 calls coming in. Would you really wait?": the most dangerous one → yes, and show why: a few weeks against a plan off by $4.6 million. Unlike those 2,000, it's a sample wired like the real fleet, read before signing. The banner carries customers meanwhile.
  • "How did you get the other managers to agree?": tests whether you moved people without authority → the firmware manager's objection, the staging and the stop you handed over.
  • "What would have happened without you?": tests whether you're inflating → the teams would have started eventually. Claim the plan and the reversal.

Red flags

  • ❌ "The supplier left us with nothing." Your own data had the 40 percent.
  • ❌ "I rebuilt the firmware and migrated the fleet." That's the teams' work in your "I", and it reads as Senior.

Principal: give executives a reset they can believe

What meets the bar

  • A funded direction overturned from outside the company, priced against the old plan in business terms.
  • Your seat is plain: you own the technical direction and make a technical call, and the CEO, CFO and board keep theirs.
  • You can say how you lost credibility, without arguing with the people who lost faith.
  • The people who built the old direction heard what survived from you.
  • A mechanism someone else runs, not a personal resolution.
  • Check your altitude: several teams aligned with no company money at stake reads as Staff; "we pivoted the business" with no technical call reads as an executive's story.

Model answer

"Page four of the notice ended our plan: no more waivers for flying beyond a pilot's sight without certified detect-and-avoid, to a standard our cameras couldn't yet meet. I'm the principal engineer for flight autonomy at a company inspecting about 60,000 line-miles a year for utilities. The direction was mine: $30 million over three years so remote operators could replace field crews, taking our cost from $620 a mile to $340, about $17 million a year. The board had funded it; $14 million was spent. Waivers were a known risk; I'd priced a tightening as years away.

That week I got remote-operations hiring paused and told the 40 engineers what we didn't yet know. Two weeks later the CTO and I took the board a reset: six months and $5 million to prove the cameras. The CEO gave that date to our three largest utilities. Regulatory affairs called six months the best case, and I showed the board only the best case. Five months on, the regulator asked for ten times the encounter data. Review time and a first-round pass were never ours to control.

I went back that week: the cameras could take two more years and might still fail. The CFO asked why anyone should believe this one. The board ordered an outside review and released the program's money by the quarter. One utility left: 5,000 miles, $4 million of revenue, $900,000 of margin.

I recommended stopping camera-only: in low sun they couldn't see a small aircraft far enough out. Our flight planning and defect models worked on field-crew flights, saving about $80 a mile. Fixed radar had an accepted path to certification, but only paid back on dense corridors, about 9,000 miles at $380. The review backed my case, and the CEO and board chose it. I told those engineers what survived, and whose roles didn't, before the paper circulated.

It cost $28 million and saves about $5 million a year after that margin, so payback went from under two years to nearly six. Any plan resting on a regulator now reaches executives as a range, each assumption we don't control and the date we'll know if it held. I wrote that page. Regulatory affairs owns it and has sent plans back."

Why this lands at Principal

  • "I showed the board only the best case": the line only a Principal would say. The notice wasn't theirs. How the board understood it was.
  • "I recommended stopping camera-only: in low sun they couldn't see a small aircraft far enough out": the engineer's call that keeps this from reading as a strategist's story.
  • "payback went from under two years to nearly six": net and unflattering. Be ready to show it: 46,000 miles at $80 and 9,000 at $240 is about $5.8 million, less $0.9 million of margin, against $17 million planned. Spend was $14M + $5M + $9M for radar and tools.

Follow-ups to expect

  • "The board lost faith in you, and 40 engineers lost their direction. What did you do in the weeks they were deciding?": the most dangerous one → don't argue for your credibility. The review and the quarterly money were right. Say what you told the engineers, what the range showed and which estimates held, e.g. radar's range was 14 to 20 months, and it took 17.
  • "What did you personally decide here?": separates the engineer from the strategist → stopping camera-only, what radar could do and where it paid. The contracts and the funding stayed with their owners.
  • "Your bet rested on waivers. Why was this a surprise?": tests whether you own the bet → what you'd priced, what the regulator had signalled, and no invented guilt.
  • "What mechanism outlived your involvement?": tests whether it outlasts you → the page, who owns it, and what it has sent back.

Red flags

  • ❌ "Nobody could have predicted that notice." A bet on waivers carried that risk. Say how you'd priced it.
  • ❌ "We renegotiated our contracts and restructured funding." That's other executives' work in your "we".

Quick check before moving on:

One story, four altitudes

SDE Senior SDE Staff Principal
Situation A yarn API retired in ten weeks Layoffs take three of six engineers, 11 weeks out Gateway supplier insolvent; 60,000 homes without data A regulator's notice stops a $30M autonomy direction
Your role Own the panel Lead the team; own the date Write the recovery plan across four teams Own the direction; CEO and board decide
The key move Rebuilt on what designers already enter Cut to 24 engineer-weeks; let the team attack it Stopped the swap; reflashed in stages Stopped camera-only; field-crew tools, radar on 9,000 miles
Result Panel kept; clicks 14 to 9 percent Six depots live; double-bookings 52 to 27 About $2.1M, against $5.4M signed and $10M projected; most homes 10–18 weeks without data $28M; payback under two years to nearly six
Reflection Read the terms; one week to save it Team hears any change within a working day Success rate from 500 real installs Regulator-dependent plans shown as ranges; regulatory affairs owns the page

Draft your answer

Step 1: Pick the right story

  • SDE: Did something outside your team take away what you built, and did you change course yourself?
  • Senior SDE: Did your team's project lose people, money or its date, and can you show the re-plan as sums?
  • Staff: Did it stop several teams, did people outside them feel it, and did a manager change a plan because of you?
  • Principal: Was a funded direction overturned, can you price old against new, and was there a technical call of yours?

Then:

  • It's from the last three years, or five for a direction.
  • You didn't mainly cause it. If you did, this is a mistake story, so save it for that question.
  • There's a number for what you kept and another for what's gone.

You don't need to have fumbled it. A recovery that went well is a strong story if you can show the first week's decisions and what the new plan gave up. Don't invent guilt to sound humble: tell a wrong turn only if it happened and you can show what changed.

If you're a new grad. A capstone whose partner pulled the dataset, a cancelled internship project or a teammate quitting the course qualify if a deadline still stood and you didn't cause it. A rejected offer or a bad grade rarely does: nothing had to be rebuilt. The director allows weeks, not quarters, but still wants the first week in order and what was gone for good. "Mid-internship, the team I was building a dashboard for was reorganized away."

Aiming a level above your current title? Tell your own-level story honestly and show you see the next one. A Senior aiming at Staff tells the party-rental story and adds: "The layoffs hit other projects that week too, and our cuts landed on each other. Next time I'd get those leads together in the first week."

Step 2: Fill the skeleton

WHAT DIED     (15s) — The plan, and what ended it, in one breath, no villain: ______
YOUR SEAT     (10s) — Your role, and what was yours to decide: ______
FIRST WEEK    (40s) — What you found out, what you stopped, who you told, in
                      order: ______
WRONG TURN    (15s) — If any, in the recovery: what you held on to, hid or rushed: ______
WHAT YOU SAVED(20s) — Kept, cut and gone for good, with who decided each: ______
WHO CAME WITH (15s) — Who lost something, what they heard from you, and when: ______
LEVEL BEAT (wherever it falls; say it once): ______
  SDE: where your part stopped, and that you raised it yourself.
  Senior: the date was yours, and so was what your team heard.
  Staff: why several teams were waiting, and who said no first.
  Principal: what you showed executives, and how they came to believe it.
RESULT        (15s) — The number against the old plan, including what's worse: ______
NOW            (5s) — What you'd do on day one next time: ______

Step 3: Calibrate it

If your answer says… It reads as… Fix
"The vendor let us down" and three more sentences on them A grievance One sentence on the cause, then your first week
"We got back on track" Too clean, or a small setback What's gone for good, with a number
Your wrong turn takes longer than what you saved A mistake story Cut it to two sentences and give the time to what you kept
A Principal answer: "The rules changed, so we pivoted" An executive's story, or Staff Your technical call, the price, and what you showed them

Step 4: Rehearse the follow-ups

Say each of these out loud once:

  1. "Walk me through the first week, in order." The one to prepare first, at every level. Days, not themes.
  2. "What's gone for good?" A number, and who lost it.
  3. "Who took this hardest, and what did they hear from you?" A role, what you said and when.
  4. "Could you have seen this coming?" What you could have known and what you couldn't, with no invented guilt.
  5. "How did your manager see it?" What they asked you, and what changed.

Common mistakes

  • Telling a missed date as a setback. If the story is mostly about the date, use the delay question.
  • Claiming a full recovery. Say what's gone before you're asked.
  • Skipping the first week. Going straight to the result hides every decision.
  • Telling a mistake as a setback. If you caused it, a follow-up will find that fast.

Key takeaways

  • A strong answer has what died, the first week in order, what you saved and what's gone, who came with you, and what you'd do sooner.
  • Any wrong turn belongs in the recovery, not in the cause.
  • As the level rises, more people have to believe the new plan.

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.