"Tell Me About a Time You Failed": How to Answer, With Examples
How to answer "Tell me about a time you failed" in a software engineering interview: what the interviewer is checking, how to pick the right failure, a STAR structure that ends on learning, and three example answers.
LiveHint.ai team · · 5 min read
"Tell me about a time you failed" makes people nervous because it feels like a trap. Admit a big failure and you look risky. Pick a tiny one and you look evasive. Say "I've never really failed" and you've answered a different question: "Are you self-aware?" The answer is no.
It isn't a trap. It's one of the most predictable questions in a behavioral round, and once you know what's being scored, it's easy to prepare.
What the interviewer is actually checking
Nobody is looking for a candidate who has never made a mistake. They're looking for evidence of four things:
| Signal | What it sounds like in a good answer |
|---|---|
| Ownership | "I made the call," not "the team decided" or "the requirements changed" |
| Judgment | You understood why it went wrong, not just that it did |
| Recovery | You limited the damage once you noticed |
| Learning | Something you do differently now, with a later example of doing it |
The last one carries the most weight. A failure followed by a concrete change in behavior is a story about growth. A failure followed by "I learned to communicate better" is a story about a failure.
How to pick the right failure
Good failure stories have a few things in common:
- It was genuinely your decision or your miss. Not a reorg, not a vendor outage, not a manager's bad call.
- It had real consequences. A slipped launch, a production incident, a feature nobody used. Something the interviewer can feel.
- It wasn't a character flaw. "I missed a deadline because I didn't break the work down" is fine. "I lied about my progress" is not.
- It's not ancient. A story from the last two or three years shows the lesson is still active.
- It's relevant to the role. For a senior role, a failure of scoping or influence lands better than a typo in a config file.
Avoid failures that are secretly strengths ("I worked too hard and burned out shipping it early"). Interviewers have heard them all and they read as dodging.
The structure: STAR, weighted toward the end
Use the STAR method, but shift the time. Most people spend 70% of the answer on the situation. For this question, the split should look more like this:
- Situation (15%): one or two sentences of context.
- Task (10%): what you were responsible for.
- Action (30%): what you did, including the mistake itself, stated plainly.
- Result and learning (45%): what happened, what you understood afterwards, what you changed, and evidence that the change stuck.
Say the failure in one clear sentence. "I shipped it without a rollback plan, and that was my call" is stronger than a paragraph of context that slowly backs into it.
Example 1: a production incident
Situation: At my last company I owned the migration of our billing service from a cron-based job to an event-driven worker.
Task: I was responsible for the design and the rollout.
Action: I tested it thoroughly in staging, but I rolled it out to 100% of customers at once instead of behind a flag. Staging didn't have the volume of retries we see in production, and on the first night we double-charged about 300 customers.
Result: I caught it from an alert within 40 minutes, rolled back, and worked with support to refund everyone the next morning. The fix itself was small: an idempotency key on the charge request.
Learning: The real lesson was about rollout, not code. Since then, anything that touches money or customer data goes out behind a flag to 1%, then 10%, with a written rollback step in the design doc. When we migrated invoicing six months later, we found a similar edge case at 1% and nobody noticed.
Notice what makes it work: a specific mistake, owned in the first person, a real cost, a fast recovery and a later example that proves the lesson stuck.
Example 2: a project nobody used
I spent about six weeks building an internal dashboard for on-call engineers because I was sure it would cut our time to diagnose incidents. I built it on my own assumptions and showed it to the team when it was done. Usage was close to zero. People already had their own queries saved and the dashboard didn't fit how they worked.
What I got wrong was skipping discovery. Now, before building anything for other people, I interview three or four of them first and show a rough version within the first week. The next tool I built that way, a runbook search, became the default starting point for incidents within a month.
Example 3: an early-career answer
If you're a new graduate, a project or internship failure works fine:
In my internship I underestimated a task to add pagination to an API. I said two days. It took two weeks because I didn't realize three clients depended on the old response shape. I didn't flag the delay until the end of the first week, which was the real mistake. My mentor had to explain it to the other team.
Since then I raise risks as soon as an estimate looks wrong, not when it's already late. In my final-year project I flagged a dependency problem on day two and we rearranged the work so nobody was blocked.
Mistakes that sink this answer
- Blaming others. Even if others contributed, focus on your part.
- Choosing something trivial. "I once forgot to update the README" tells the interviewer you won't open up.
- Ending on the failure. Always land on what you changed.
- Vague learning. "I learned the importance of communication" is not a lesson. "I now post a status update every Friday on any project longer than two weeks" is.
- Going too long. Aim for about two minutes. See how long interview answers should be.
Expect follow-up questions
Interviewers usually dig in:
- "What would you do differently if it happened today?"
- "How did your manager react?"
- "Did anyone else see the problem coming?"
- "How do you know the change actually worked?"
Prepare one or two sentences for each. The follow-ups are where a rehearsed story either holds up or falls apart.
Practice it out loud
Write your story down in STAR form, then say it out loud and time it. Run it in an AI mock interview set to a behavioral round, where the interviewer asks follow-ups and the scorecard shows whether you spent too long on context. If you're also preparing related questions, see our guides to your greatest weakness and conflict with a coworker.
Keep reading
13 min read
STAR Method Examples: 12 Interview Answers That Work
Twelve full STAR method examples for software engineers, from conflict and failure to tight deadlines and production incidents, plus a checklist to adapt them.
- behavioral interview
- STAR method
- interview tips
- software engineering
6 min read
The STAR Method: How to Answer Behavioral Interview Questions (With Examples)
A practical guide to the STAR method for behavioral interviews: what interviewers listen for, how to structure each part, and full example answers for engineers.
- behavioral interview
- STAR method
- interview tips
- software engineering