What this question is really asking
The trap in this question is that it feels like a question about failure. It is not.
It is a question about accountability — whether you can describe something going wrong, name your own part in it plainly, and show what changed as a result.
Which is why the two instinctive answers both fail. The humblebrag — "I took on too much because I care so much" — dodges the question, and interviewers simply ask again. The genuinely catastrophic story creates a new problem you cannot resolve in the room.
What you want is a real failure, with real consequences, that you owned and resolved.
How to structure your answer
Choosing the story is most of the work
Before thinking about how to tell it, spend your time choosing well. A well-chosen story almost tells itself.
A usable failure meets four tests:
It was genuinely your responsibility. Not something that happened near you. Something where your decision or your omission caused the problem.
It had a visible consequence. Someone was affected. A deadline slipped. Work had to be redone. A client noticed. If nothing happened, it was not a failure.
It is finished. Resolved, in the past, with a clear ending. An ongoing situation leaves the interviewer with an open question.
It produced a lesson you can evidence. Something changed afterwards, and you can point to it.
Two things to avoid entirely: anything involving dishonesty or an ethical lapse, and anything so severe it raises real doubts about your judgment.
The structure, and the ratio that matters
Most people know the STAR structure — Situation, Task, Action, Result. That is fine, but for this question the ratio matters more than the labels.
Aim for roughly:
- 20 percent setting the scene
- 30 percent what went wrong and your part in it
- 50 percent what you did about it and what changed
Most candidates invert this completely, spending two minutes on background and fifteen seconds on reflection. The second half is what is being assessed.
Naming your part
This is the sentence that determines how the whole answer lands.
Weak: "There were some miscommunications on the team and the deadline slipped."
Passive, vague, and it distributes the blame. Interviewers hear this constantly and know exactly what it means.
Strong: "I assumed the data had already been checked by someone else and I did not verify it myself. That is what caused it."
Direct, specific, first person. That is what accountability sounds like, and the difference is audible immediately.
Watch out for passive voice generally. "Mistakes were made" and "the deadline was missed" hide the exact thing the question is asking you to reveal.
Describing what you did next
Failure is universal. The response is not, and this is the part they actually care about.
- Did you flag it immediately, or hope it would resolve itself?
- Did you fix it yourself, or wait to be told?
- Did you tell someone senior before they found out from someone else?
Weak: "I learned to be more careful."
Strong: "I flagged it to my manager the same afternoon rather than waiting for the client review, and I rebuilt the analysis that evening so we could go into the meeting with the corrected version."
The second version describes behaviour under pressure, which is what they are trying to predict.
The lesson must be specific
Generic lessons signal a rehearsed answer.
Weak: "I learned the importance of communication."
Everyone says this. It means nothing.
Strong: "Now, whenever I build on someone else's work, I confirm in writing what I am assuming has already been checked. It takes two minutes and it has caught two errors since."
That is a behavioural change, not a sentiment.
Close with evidence it stuck
One sentence about a later situation where you did it differently.
"On the next project I did exactly that at the start, and it caught a problem with the source file before we had built anything on top of it."
This is what turns a story into a demonstrated change. Almost nobody includes it, and it is the strongest single thing you can add.
A full worked example
"In my second year I led the analysis for a group consulting project with an external client. We were modelling their cost base, and I built the model using a dataset one of my teammates had pulled together.
I assumed the data had been cleaned. It had not — there was a duplication issue affecting about 15 percent of the rows. I did not check it myself, and I built two weeks of analysis on top of it. We found out three days before the client presentation, when the numbers did not reconcile with something the client mentioned.
I flagged it to the team and to our faculty supervisor the same afternoon rather than trying to quietly fix it. Then I rebuilt the model over two evenings, and we presented with corrected numbers — a day later than planned, which the client accepted, but it was avoidable and it was my fault.
What changed is that I now confirm explicitly, in writing, what has and has not been checked before I build on someone else's work. It takes two minutes.
On my internship last summer I did exactly that on the first day of a live process, and it surfaced a mismatch in the source data before anyone had built anything on it."
About ninety seconds. Real consequence, clearly owned, specific response, evidenced change.
Tone
Matter-of-fact. Something went wrong, you were responsible, here is what you did.
Two failure modes on tone. Deflecting reads as someone who will blame colleagues. Over-apologising reads as fragile, and shifts the interviewer from assessing your judgment to worrying about your resilience.
One acknowledgement is enough. Then move to what you did.
What the interviewer is really checking
Can you be accountable without falling apart?
Two failure modes get flagged: deflecting blame, and over-apologising to the point of seeming fragile. Matter-of-fact ownership is the target.
Is your calibration sensible?
Something trivial suggests you either have not failed at anything meaningful or do not recognise when you have. Something catastrophic raises a different concern.
Is the lesson real?
Generic lessons signal a rehearsed answer. Specific behavioural changes signal actual reflection.
Are you honest?
This is one of the more reliable honesty tests in a behavioural interview. A manufactured failure tends to be suspiciously well-shaped, and follow-up questions expose it quickly.
How do you talk about other people?
Someone who blames a teammate in an interview will blame a colleague on a deal. Even when others genuinely contributed, keep the focus on your part.
Frequently asked
How big should the failure be?
Big enough to have had a visible consequence — someone was affected, a deadline slipped, work had to be redone — but not so big that it raises doubts about your judgment. A project that went wrong on your watch is the right size.
Can I use a group project?
Yes, and many strong answers do. The key is that your part in the failure must be clear and owned. A group project where your point is that others let you down is the version that does not work.
What if the failure genuinely was not mostly my fault?
Then choose a different story. Interviewers are not assessing whether the failure was fair; they are assessing whether you can own your part in something. A story where your part was minimal does not give you anything to demonstrate.
Can I reuse this story for greatest weakness?
Avoid it. Interviewers often ask both, sometimes in the same interview, and reusing one story makes it look like you have only one example. Prepare distinct stories for failure, weakness, conflict and mistake.
What if I have not really failed at anything?
You have — the difficulty is usually recognition rather than absence. Look for a time you misjudged a timeline, assumed something you should have checked, or committed to more than you could deliver. Those are the everyday failures interviewers expect to hear.
Common mistakes
The humblebrag.
"I am a perfectionist." "I worked too hard." "I cared too much about the outcome." These are non-answers, and interviewers will simply ask again with less patience.
Blaming the team, the manager or the circumstances.
Even when partly true, it is the wrong emphasis. Mention context in one clause, then return to your own role.
Picking something trivial.
A missed reading for a seminar is not a failure. No consequence means no answer.
Picking something unresolved or ethical.
Anything involving dishonesty, or a situation still ongoing, creates concerns you cannot close in the room.
Spending the whole answer on setup.
Two minutes of background and fifteen seconds of reflection is the most common structural error here.
A cliché lesson.
"I learned the importance of communication" tells them nothing about you.
Over-apologising.
One acknowledgement is enough. Repeated self-criticism shifts the assessment from your judgment to your resilience.
Passive voice.
"Mistakes were made" hides exactly what the question is asking you to show.
Ready to practice?
Answer this question out loud with an AI interviewer.