Why structured interview questions predict job performance better
The single most consistent finding in 100 years of industrial-organisational psychology research is this: structured interviews — where every candidate is asked the same questions, scored against the same rubric — predict job performance roughly twice as well as unstructured interviews where the interviewer "has a chat" and forms an impression. The unstructured interview is one of the worst predictors of job performance ever measured. The structured interview is one of the best.
The four question types that work
- Behavioural questions ("tell me about a time when...") — past behaviour as a predictor of future behaviour. Strong evidence base. Use the STAR framework (Situation, Task, Action, Result) to structure your scoring.
- Situational questions ("what would you do if...") — hypothetical scenarios that test judgment and approach. Useful when candidates lack direct prior experience.
- Role-specific / technical questions — direct assessment of the skills the role requires. Should be calibrated to the seniority level, not theoretical.
- Motivation questions — understanding why this candidate, this role, this company, this moment. Predicts retention more than performance.
The questions that don't work
- Brain teasers and trick questions ("how many tennis balls fit in a 747?") — Google retired these years ago after their own data showed zero predictive value.
- "Tell me about yourself" as a content question — fine as warm-up, but the answer doesn't predict performance. Most candidates rehearse it; you learn what they want you to learn.
- Hypothetical "What's your greatest weakness?" — coached answer space. The honest answers don't reach you.
- Anything that invites unconscious bias — questions about hobbies, family, weekend plans, "where you're really from". These create rapport-driven hiring decisions instead of evidence-driven ones.
Score immediately, score blindly. Score each candidate's answer on a scorecard within 24 hours of the interview, before reading other panel members' scores. Aggregate scores at the end. This single change — score before discussing — substantially reduces "anchoring" bias where the loudest panel member shapes everyone else's view.
How many questions should one interview cover?
- 30-minute screen: 4–6 questions plus 5 minutes for candidate questions.
- 45-minute hiring manager: 6–8 questions plus 10 minutes for candidate questions.
- 60-minute panel: 8–10 questions plus 15 minutes for candidate questions.
- Technical interview: 1–2 deep problems, not 8 shallow ones.