Most "behavioural questions" lists online run to 50, 100, 200 entries. Useless. Nobody asks 200 of these in a real round. There are about 25 patterns that account for most behavioural questions at tech companies in 2026, and most actual interviews pick five of them.

This is that list. Below each question, what the panel is actually testing. The STAR framework still works for the answer structure, but the structure isn't the interesting part. The interesting part is which stories you've prepared, and how specifically.

The STAR framework, briefly

You've read this 30 times. Here it is in 50 words: Situation (where, when, who), Task (what was your role specifically), Action (what you did, with specifics), Result (what changed, with a number if possible). The big mistake people make is spending too long on S+T and skipping the specifics in A. Flip the ratio.

Stories you should prepare in advance

Most behavioural questions are testing one of seven dimensions. Prepare one strong story per dimension and you can answer 80% of what comes up.

Pick one story per dimension. Write it down. Time it at 90 seconds spoken. Most of your behavioural prep is in having the seven stories ready, not in memorising answers to 200 questions.

The 25 questions worth knowing

Grouped by dimension. Each question maps to one of the seven stories above.

Conflict / disagreement

1. Tell me about a time you disagreed with your manager.
2. Describe a conflict with a teammate. How did you resolve it?
3. Have you had to push back on a stakeholder's request?
4. When did you have to deliver feedback someone didn't want to hear?

Failure

5. Tell me about a project that failed.
6. Describe a time you missed a deadline.
7. What's a mistake you made that you regret?
8. When did you make a decision that you'd undo if you could?

Leadership without authority

9. Tell me about a time you led without a formal title.
10. Describe a time you influenced a peer to change their approach.
11. How do you motivate teammates who report elsewhere?

Ambiguity

12. Tell me about a project where the requirements weren't clear.
13. What did you do when you didn't know what to build next?
14. Describe a time you had to make a decision with incomplete information.

Trade-off

15. Tell me about a time you had to choose between speed and quality.
16. Describe a technical decision where you had to pick between two reasonable options.
17. When did you have to scope something down for a deadline?

Mistake recovery

18. Tell me about a production incident you caused.
19. How did you handle a bug that made it to customers?
20. Describe a time you broke something during a deploy.

High-impact shipping

21. What's the most impactful project you've shipped?
22. Tell me about a project where you moved a specific metric.
23. Describe something you shipped that you're proud of.
24. What's a project you took ownership of end-to-end?
25. Tell me about a time you shipped under tight constraints.

The four classic openers

These don't fit the STAR framework cleanly because they're not story questions. Each has its own structure. Worth dedicated prep:

• "Tell me about yourself" : 90-second past/present/why-here structure
• "What's your biggest weakness?" : three-part: weakness, what you've done, evidence
• "Why should we hire you?" : value-anchor: their problem, your win, the bridge
• "Where do you see yourself in 5 years?" : skill direction + growth path + implicit commitment

What we hear from mock rounds

Across the behavioural mocks LastRound AI's copilot runs, four failure modes account for most of the weak-or-fail signals:

1. Recycled stories. Candidates use the same project for three different questions. Panels catch this. Pick seven different ones if you can.
2. "We" instead of "I". The question is about you. Saying "we built X" doesn't tell the panel what you specifically did. Be willing to say "I".
3. No numbers. "It worked well" isn't a result. "We cut latency from 800ms to 200ms" is. If you don't have a metric, at least name the visible outcome.
4. Skipping the reflection. Most questions want a beat at the end where you say what you'd do differently. Without it, the answer feels like a brag.

The reference panels use as the de-facto standard for behavioural rubrics is Amazon's Leadership Principles. Even non-Amazon companies grade against something shaped like LPs (ownership, dive deep, deliver results, earn trust). If you've never read them straight through, spend ten minutes on the page.