Learn from a failure
The first sections are free. The sample answer, follow-up questions and quiz are part of Pro.
What it tests
Whether you can own an outcome that went wrong, say plainly which part of it was yours, and show that it changed how you work. The prompt has two halves, and the second one carries most of the weight: what you did next, once the failure was clear, and what you did differently the next time a similar situation came up. Amazon's recruiters publish advice to come prepared with examples of taking risks, failing and growing, and one of its recruiting leaders is quoted there saying failure is fine as long as you improve from what it taught you. One engineering interview guide lists what a failure story shows an interviewer: whether you can say accurately what went wrong, whether you own your part or blame circumstances, how you recovered, whether you're candid about real difficulty, and whether you've applied the lesson since. A strong answer gives all five without being asked.
What the interviewer listens for
| What | Why it matters |
|---|---|
Situation | The project, who depended on it, and what a miss would cost them, in a sentence or two. If the interviewer can't tell what was at stake, they can't tell whether your response matched the size of the problem, and a failure with no stakes reads as a safe pick rather than a real one. |
Task | The prompt says "something you were responsible for," so name the responsibility exactly: the deliverable, the date, the system, the decision that was yours to make. This is the line the rest of the story is measured against, and it's what keeps the answer from drifting into a team failure you merely watched. |
Action | Two sets of steps. First, what you did in the hours and days after the failure showed: who you told, what you contained, what you fixed. Second, what you changed afterward. Both need to be things you did, in order. "We did a retro and moved on" gives the interviewer nothing to write down. |
Result | Two outcomes, stated honestly. The cost of the failure itself (the delay, the customers affected, the trust lost), and the outcome of your response. Softening the first makes the second look smaller than it was. |
Ownership | Which part of the failure was yours. Most real failures have several causes; naming the one you controlled, in the first person, is what separates owning it from describing it. Amazon's own interview advice asks for "I" when you describe your part so the interviewer can see your individual role, and that applies to the failure as much as to the fix. |
Decision | The decision that led to the failure, with the reasoning you had at the time, or the decision you made in response. Saying what you believed when you made the call shows the failure came from a judgment that turned out wrong, and a judgment is something you can examine and change, which gives the lesson somewhere to land. |
Trade-off | What you were trading when you made the call: the date against the extra test, the quick fix against the full one. Naming the other option and why you passed on it shows you understood the choice, and it sets up the lesson, which can then be about how you'd weigh that trade next time or who you'd tell that it was being made. |
Learning | A lesson at the level of behavior, plus a later moment when you acted on it. "I learned to communicate better" is a sentiment. "I now tell stakeholders the day an estimate slips, and here's the project where I did" is evidence that the lesson stuck. |
Specifics | One real failure, with the system, the date, the people affected and what you said to them. A general claim about how you handle setbacks can't be checked with a follow-up; a specific story can, and the follow-ups are where the depth shows. |
Numbers | Optional, but useful in two places: the size of the miss (how many weeks late, how many customers, how long the outage) and the outcome of the later project where you applied the lesson. The same Amazon advice asks for quantifiable detail where it exists and accepts specific anecdotal feedback where it doesn't. |
How to structure the answer
| Part | What to say |
|---|---|
Responsibility and stakes | What you owned, who depended on it, and what a miss would cost them. One or two sentences, then move on. |
The miss | What went wrong, stated plainly and without softening. Say what you did or decided that contributed, and what you believed at the time that made it seem reasonable. |
What you did next | The response in order: who you told and when, what you contained or rolled back, what you fixed, and how you wrote it up afterward. This is where the second half of the prompt gets answered. |
Cost | What the failure cost, honestly: the delay, the customers affected, the trust that took time to rebuild. |
Lesson, then proof | The behavior you changed, and a later project where you acted on it and it mattered. Without the later example the lesson is a claim; with it, it's a pattern. |
Pitfalls
| Pitfall | Why it hurts |
|---|---|
Disguising a success as a failure. | "We shipped late but it was the best launch we ever had" tells the interviewer you'd rather not talk about a real miss. One interviewer who writes about asking the weakness question lists a "disguised strength" as a red flag for that question. Our view: the same logic applies here, because dodging the question is itself the signal. |
Offering a lesson without showing a change in behavior. | "I learned the importance of testing" could be said by anyone, including someone who still skips tests. The third thing a strong answer does is show a later situation where you acted differently because of this one. If you don't have that yet, say what you've changed in how you work and give the most recent instance of it. |
Spending the time on why it wasn't your fault. | Even when the context was genuinely hard, a story that centers on the manager, the vendor or the deadline tells the interviewer where your attention goes when things break. Blaming the team or the manager reads as poor accountability. Give the context in a sentence, then spend the rest on the part you controlled. |
Picking a failure with no consequence. | A miss that nobody felt gives you little to work with: the response is short, and the lesson has nothing to attach to. A typo in a pull request is too small to be worth telling. Choose a miss that someone else felt: a delayed team, an affected customer, a number that moved the wrong way. |
Stopping at the confession. | "I shipped a bad design and we're still living with it" answers the first half of the prompt and ignores the second. The question asks what you did next; if the honest answer is that you didn't do much at the time, say what you've done since to repair or contain it. |
Still defending the call. | "I'd make the same decision again" can be true, but if it's the whole reflection, it sounds like the lesson wasn't learned. If the decision was reasonable on the information you had, say so, then say what you'd do to get better information sooner next time. |
Reaching back to school or an early internship. | A failure from years ago shows how you behaved years ago. Use a recent, professional one where you can. The exception is a junior candidate with little work history, where a class project or internship story is fine as long as the response and the lesson are recent and real. |