How do you answer behavioural interview questions?

A behavioural question starts "tell me about a time when" and wants one specific past event. Answer it as Situation, Task, Action, Result, and spend most of the answer on Action and Result — where most people underinvest. A model cannot supply the story, and an invented example collapses under one follow-up question.

• The tell is the tense: a behavioural question asks what you did, a hypothetical asks what you would do, and they want different answers
• Situation and Task belong in the first two sentences; a minute in with no Action yet means the answer is lopsided
• Six to eight stories cover a typical interview, because one project that went badly answers four separate questions
• "I haven't faced that exactly, but the nearest thing was…" is acceptable; a fabricated example is not, and one follow-up usually exposes it
• FluxResume's Job Based Resume analysis puts a View AI Reasoning & Interview Prep expander on each proposed edit, showing what follow-up that wording invites

How do you recognise a behavioural question?​

By the tense. A behavioural question asks about something that already happened — "tell me about a time when", "describe a situation where", "give me an example of", "walk me through how you handled" — and the answer has to be one specific event with a date attached in your head.

Two look similar and are not. "How would you handle a team member who missed deadlines?" is hypothetical and wants your reasoning; "what would you do in your first ninety days?" is a plan question. Either allows a story as evidence, but neither is satisfied by one. Two others are disguised: "what is your greatest weakness" and "why did you leave that job" are behavioural, because the only convincing answer is a specific instance and what you did about it.

What does STAR actually require?​

It requires four parts in a fixed order, with the time distributed nothing like the way most people distribute it.

• Situation — one or two sentences. Where you were, what the constraint was.
• Task — one sentence. What you specifically owned, as distinct from what the team owned.
• Action — the bulk of the answer, in the first person singular. An interviewer cannot assess a contribution they cannot locate, and an answer in the first person plural describes a team you watched.
• Result — what changed, stated so the interviewer can check it with a follow-up.

Here is the correction worth taking seriously, because it is the most common failure in a behavioural answer: almost everybody over-invests in Situation and under-invests in Result. People spend ninety seconds on background — the org chart, the project history, the previous vendor — then arrive at "so I fixed it" and stop, leaving the interviewer with a well-described problem and no evidence about you.

Two mechanical fixes deal with most of it. Cap the setup at two sentences and let them ask for more context, which they will, then say the Result out loud as your last sentence. Where there is no number — plenty of real work produces none — a before and after works. A line like "reduced onboarding from three weeks to eight days" is the shape to aim for, and if the honest version is "the process survived two staff changes without me", say that rather than inventing a percentage.

How do you build a story bank instead of one answer per question?​

Stop mapping one story to one question. Six to eight chosen for range cover a typical interview several times over, because most behavioural questions ask about the same handful of situations from different angles. Pick ones that each carry several themes:

• A project that went wrong — failure, pressure, accountability, what you would do differently
• A disagreement with someone you kept working with — conflict, influence, difficult feedback
• Something shipped under a hard constraint — prioritisation, ambiguity, limited resources
• A time you changed your mind on evidence — learning, decisions under uncertainty
• Something you led without being anyone's manager — leadership, initiative, mentoring
• A time you said no to a deadline — judgement, managing expectations upward

Write each in the four-part shape once, then leave it alone. In the room you are choosing an emphasis rather than retrieving a paragraph: the same failed migration is a failure story, a stakeholder story or a prioritisation story depending on which part you expand.

The raw material is already written down: the achievements you condensed to fit the page — the ones behind a summary of qualifications — are these events minus the context that makes them interesting. Reread the resume you submitted before you build the bank, so the two agree; everything else before the day is in how to prepare for a job interview.

What do you say when you have no example?​

Say so, and offer the nearest real thing in the same breath: "I haven't managed a team that size, but the closest is coordinating four contractors across two time zones for a quarter."

That answer holds up better than people expect, for a reason unrelated to honesty being nice. An interviewer's next move after any story is a follow-up, and a fabricated example has no second layer. Asked what the other person said, what the numbers were before, or why you chose that approach, the invented story contradicts itself or turns vague at the moment specificity is being tested. Nobody has to catch a lie; they only notice one answer is thinner than the others.

Two substitutes are legitimate: an adjacent situation with the difference stated plainly, or reasoning flagged as reasoning. What you cannot do is claim someone else's work, or move a story from a job where you watched it to one where you led it. If the gap is structural rather than situational, that is a factual sentence rather than a story, and how to explain an employment gap covers what belongs in it.

How do you handle the failure and conflict questions?​

These two carry the most weight and get the weakest answers, because people try to make them not count.

On failure, choose a real one with real consequences and put most of the answer after the failure rather than before it: what went wrong, what it cost, what you changed, and what evidence exists that the change held. The answers that fail are a failure that was somebody else's, one that was trivial, and one with no change at the end. "I'm a perfectionist" is not an answer.

Conflict questions are checking whether you can disagree with someone and still work with them next week, so the answer needs the other person to be reasonable. Describe their position as they would describe it, say what was driving it, and say what you did. You do not have to have won: "we went with their approach and I was wrong about the risk" is a strong answer. Any version where the other person is stupid or political tells the interviewer what you will say about your new colleagues.

Can an AI assistant help you drill this?​

For structure, yes. For content, no. A model can read a story you wrote and name which of the four parts is thin, which is the one failure you cannot see in your own writing.

Here is a rough draft of an interview story, written as facts, not polished:

[paste your story — where you were, the constraint, what you did, what happened]

Do four things, in this order:
1. Split what I wrote into Situation, Task, Action, Result. Quote my own words.
2. Tell me which of the four is missing or thinnest, and why.
3. List the follow-up questions an interviewer would ask that my draft
   does not currently answer.
4. Tell me what is unverifiable as written — anything where I claimed a
   result without saying how it was measured.

Do not rewrite the story. Do not add any detail I did not give you.

Those last two lines matter. Without them you get back a polished paragraph containing facts you never supplied, and rehearsing it commits you to details you cannot defend.

The second scaffold predicts rather than critiques, which is the better use of the tool:

Here is a job posting:

[paste the entire posting]

Generate the 12 behavioural questions this employer is most likely to ask,
based only on what the posting states. For each one:
- quote the line in the posting it comes from
- name the competency it is testing, in one phrase

Group them so I can see which questions one story would answer.
Do not write answers.

Run that, then map your stories onto the list. Where three questions collapse onto one story you are covered; where a competency the posting repeats has no story against it, you have found the thing to prepare — or to be honest about.

Inside FluxResume, AI Chat runs the same drill from an Interview Prep chip with the posting and your submitted resume already attached — how that attachment works is covered in how to prepare for a job interview. Public entry point: https://fluxresume.io/career-prep/.

What none of it does is supply the story. There is no story bank feature, nothing records practice answers, and no tool can tell you whether an event happened the way you remember it. The eight events are yours, and they either exist or they do not.

Questions people ask about behavioural interviews​

How long should a behavioural answer be?​

Ninety seconds to two minutes, with the setup inside the first fifteen seconds. Past two minutes an interviewer stops tracking the thread and waits for a gap. Finish on the result rather than trailing off, and let follow-ups carry the detail — an answer inviting three follow-ups covers more ground than a monologue.

Can you use the same story twice in one interview?​

With different interviewers, yes, and it is normal in a loop. With the same interviewer, no — reusing it signals one experience rather than several. Say "I'll use a different project for this one" and switch. That is the argument for six to eight rather than three: four conversations in a day will otherwise have you repeating your strongest example until it stops landing.

Should you say "Situation" and "Task" out loud?​

No. STAR is a preparation structure rather than a script to narrate, and announcing the labels makes a rehearsed answer sound more rehearsed. Tell the story in that order and leave the scaffolding out. The one exception is a short "the result was" before your last sentence, which stops you ending on the action and leaving the outcome implied.