The STAR Method for Tech Interviews: Examples by Level

Most engineers can recite the STAR method: Situation, Task, Action, Result. Far fewer pass the behavioural round with it, because the acronym says how to order an answer and nothing about what the answer has to prove. A senior candidate who tells a tidy four-part story about fixing a bug has followed the format and still answered at the wrong level.
This guide covers how much time each part gets, what has to change in the story as the level goes up, three worked answers (mid, senior and staff), the follow-up questions that decide the round, and the cases where STAR is the wrong tool.
The four parts and how long each should run
- Situation. The context: where you were, what was going on, why it mattered. Two or three sentences.
- Task. What you personally were on the hook for. One sentence.
- Action. What you did, in the first person, including the decisions you made and the options you turned down.
- Result. What changed, with a number where one exists, and one line on what you would do differently.
The usual failure is a Situation that eats the answer. A workable split is about 15 percent Situation, 10 percent Task, 50 percent Action and 25 percent Result. Spoken, the whole thing should run 90 seconds to two minutes. Shorter and there is nothing to probe; longer and the interviewer starts planning the interruption.
One change to the textbook version is worth making: say the headline first. "This is about migrating checkout off a payment API with a 60-day deadline" tells the interviewer which story they are about to hear and lets them redirect you before you spend two minutes on the wrong one.
What the level changes
The structure stays the same from new graduate to principal. What the interviewer listens for inside it does not.
| Level | The Situation is about | The Action has to show | The Result is measured in |
|---|---|---|---|
| Mid | A defined problem inside your team | Sound execution, one or two real decisions, asking for help well | The feature, the deadline, the metric you owned |
| Senior | A problem nobody framed for you, often two teams | Trade-offs, a disagreement handled, other people brought along | A team outcome and something that outlasted it |
| Staff | An ambiguous problem across several teams | Choosing what to work on, influence without authority, saying no | An organisation outcome, including work you stopped |
A story cannot be promoted by adjectives: calling a project "critical" does not make it senior. The level sits in which decisions you put in the Action. If you led the design and also wrote most of the code, the senior version spends its time on the design choice and the people, and one sentence on the code.
A worked answer at mid level
The three answers below are invented examples, written to show shape. Do not borrow them; interviewers ask follow-ups that only the person who lived the story can answer.
Question. "Tell me about a time you delivered under a hard deadline."
Headline. "I moved our checkout off a payment API that was being shut down, with 60 days' notice."
Situation. "Our payment provider announced it was retiring the API version we used. If we missed the date, checkout would fail for the markets still on it, which was a large share of our orders."
Task. "I owned the migration: scoping it, planning the rollout and getting us off before the cutoff."
Action. "In the first week I audited every call site and found that only the checkout path carried real risk. I proposed putting both API versions behind a small adapter so we could switch market by market with a flag, not all at once. My tech lead worried the adapter was extra work; I showed that it cost about three days and removed the need for a weekend cutover. I built it, moved our two smallest markets first, and used the errors we saw there to fix the rollout before touching the large ones."
Result. "We finished nine days early with no checkout downtime. The staged rollout caught two bugs that would have hit our biggest market. If I did it again I would have asked the provider for a sandbox with production-like data in week one; we lost a few days to test data that did not match."
Why it reads as mid level. The problem arrived already defined, the scope is one team, and the judgement on show is good execution with one decision defended. Right for the level, thin for a senior loop.
A worked answer at senior level
Question. "Tell me about a time you disagreed with a technical decision."
Headline. "I argued against a full rewrite of our indexing pipeline and got agreement on a narrower fix."
Situation. "Search results were going stale for hours after catalogue updates. The team that owned indexing proposed a rewrite on a streaming stack, estimated at two quarters. My team consumed the index and felt the staleness most."
Task. "Nobody had asked me to weigh in. I thought the rewrite solved a wider problem than the one we had, and I took it on myself to test that before the plan was approved."
Action. "I spent three days tracing where the delay came from and found that most of it sat in one batch step that waited for a full export. I wrote that up in two pages with the measurements and sent it to the indexing lead before the review, so it would not land as a surprise in front of his team. He disagreed at first: the rewrite also paid down debt he cared about. We agreed on what we were deciding, which was freshness this quarter, and I proposed making that one step incremental, with a clear list of what it would not fix. I offered one of my engineers for four weeks to help. When our director asked for a recommendation I presented both options and his objections as well as mine."
Result. "Staleness dropped from hours to a few minutes in six weeks. The rewrite went ahead the following year with a smaller scope, and the indexing lead used my trace as its baseline. What I would change: I should have talked to him before I started measuring. He heard about it second-hand, and that cost a week of trust."
Why it reads as senior. Nobody assigned the problem. The Action is mostly about a disagreement with a peer on another team, and the answer shows the other side's reasoning fairly. The Result includes something that lasted and an honest cost.
A worked answer at staff level
Question. "Tell me about the most ambiguous problem you have owned."
Headline. "On-call was burning out four teams, no one owned the cause, and I turned it into a funded programme and stopped one project to pay for it."
Situation. "Four teams shared a platform and each ran its own on-call. Pages had roughly doubled in a year, two people had asked to leave the rotation, and every team's retro blamed a different cause."
Task. "It was nobody's job. My director asked me to find out whether this was one problem or four, and to come back with a recommendation."
Action. "I read three months of pages and classified them myself before talking to anyone, because each team had a theory. About half traced back to one shared dependency with no owner and no service level objective. I wrote a short document that separated what I knew from what I suspected, and took it to the four tech leads one at a time before any group meeting. Two agreed quickly. One objected that fixing the dependency would delay a launch his team had promised. I did not have the authority to move his roadmap, so I laid out the cost of both paths for our director with his numbers, and recommended pausing a migration my own team wanted, to free two engineers. I set a target, pages per week per rotation, and a date when we would review whether it was working."
Result. "Pages fell by about half over a quarter and the dependency now has an owning team. The launch slipped three weeks, which I had flagged in advance. The migration I paused is still not done, and I think that was the right call. What I would do differently: set the review date in the first document. We argued about success criteria later than we needed to."
Why it reads as staff. The work began before the problem had a name. Most of the Action is deciding what to work on and moving people who do not report to you, and the Result admits a trade-off the candidate chose. There is almost no code in it. The staff engineer interview guide covers the rest of that loop.
The follow-up questions that decide the round
The first answer is the opening move. Trained interviewers spend more time on what comes after it, and each follow-up tests something specific.
- "What exactly did you do, as opposed to the team?" Tests ownership. Have one sentence ready that draws the line honestly.
- "What other options did you consider?" Tests judgement. Name the one you rejected and the reason.
- "How did you know it worked?" Tests whether the result is real. Know where the number came from and what it was before.
- "Who disagreed, and what did you do about it?" Tests how you work with people. "Nobody disagreed" is rarely believed.
- "What would you do differently?" Tests self-awareness. Give a specific change, not a virtue dressed as a flaw.
- "What happened afterwards?" Tests whether the work lasted. Know the state of the thing six months on.
Answer follow-ups in 20 to 30 seconds. A second two-minute story in reply to "why that option?" reads as evasion. For each story, prepare two layers of detail you did not use in the main answer: one technical, one about people.
This is the part you cannot practise alone, because you already know which questions you hope will not come. It needs a second person who interrupts, ideally a behavioural interview coach or a colleague who has run loops.
When STAR fails
STAR fits "tell me about a time" questions. It breaks, or needs bending, in these cases.
- Hypothetical questions. "How would you handle a teammate who misses deadlines?" asks for your approach. Give it in three or four sentences, then anchor it: "The last time this happened, I did this." A full STAR story with no stated approach misses the question.
- Questions that are not about an event. "Tell me about yourself," "why this company" and "what are you looking for" need a short structured pitch, not a Situation.
- Failure stories. The format pulls you toward a happy Result, so people pick failures that were secretly successes. Interviewers notice. Tell a real failure, keep the Result honest, and spend the extra time on what you changed afterwards and the evidence that it stuck.
- Long, multi-phase work. A year-long staff project does not fit one arc. Give the headline and the outcome first, then two or three decisions, each a small Action and Result of its own, and offer to go deeper on whichever one the interviewer picks.
- Interviewers who interrupt. Some cut in within twenty seconds, and that is usually interest. Follow them. You need to be able to enter your story at any point, which a memorised script does not allow, so rehearse the facts and the order, not the sentences.
Variants and the one addition worth keeping
SOAR, CARL and STARL rearrange the same idea. The addition worth keeping is the L for Learning: one sentence of reflection at the end of the Result answers "what would you do differently?" before it is asked. For Amazon, each story also needs mapping to the leadership principles; one strong story often covers two or three.
A two-week practice routine
Twenty minutes a day is enough, if it is out loud.
- Days 1 and 2. List eight to ten stories as bullets, four lines each, with a headline. The story bank guide has the slots to fill.
- Days 3 to 5. Record two stories a day on your phone with a timer. Listen once and note three things: how long the Situation ran, how often you said "we", and whether the Result had a number.
- Days 6 and 7. Write three likely follow-ups under each story and answer them aloud in 30 seconds each.
- Days 8 to 10. Draw questions at random from a list for your target company and pick the story on the spot. Choosing under pressure is a separate skill from telling.
- Days 11 and 12. One session with another person who interrupts. Ask them to write down every point where they lost the thread.
- Day 13. Cut. Any story that still runs past two minutes loses its weakest detail.
- Day 14. Headlines only. Read the list once and stop.
Where coaching fits
Solo practice fixes length and structure. It does not tell you whether your senior story sounds senior to someone who has graded senior loops, and it cannot ask the follow-up you were hoping to avoid. As of November 2026, 27 of the 40 coaches on techinterview.coach list behavioural interviews as a specialty; a session with one of these behavioural interview coaches can cover your two weakest stories, the follow-ups, and calibration against the level you are applying for. Before paying anyone, run the free AI practice interview on the behavioural round (free, sign in to start) at the practice page and read the written recap.
Related guides
- Building your story bank: the raw material STAR formats, with twelve slots to fill.
- Amazon leadership principles examples: sixteen STAR stories mapped to principles.
- Meta behavioral interview questions: the same technique against a signal-based rubric.
- Mock interviews: how to run the practice session in step 5.
- System design questions by level: the same level expectations in the design round.