Interview Science8 min read
First-round interview questions that actually predict job performance
"Tell me about yourself." "Where do you see yourself in 5 years?" "What's your greatest weakness?"
These questions are asked millions of times per day in first-round interviews. They are also among the least predictive questions you can ask.
This article is about what research — and practitioners running thousands of interviews — says actually works.
Why most first-round questions fail
Schmidt and Hunter's foundational meta-analysis of 85 years of selection research found that the average unstructured interview has a predictive validity of about 0.38 — meaning it explains about 14% of variance in job performance. That's not nothing, but it's low enough that you're making a lot of mistakes.
The main culprit: most interview questions measure interview performance, not job performance. They test how well someone has prepared for interviews and how articulate they are under a specific social pressure — not how they'll actually think and work.
Structured interviews — where every candidate is asked the same questions, in the same order, scored against the same criteria — show validity of 0.51 or higher. The structure is the intervention.
Question types that predict performance
High predictive validity
Behavioural (past-based) questions
The STAR framework (Situation, Task, Action, Result) works because past behaviour is the best predictor of future behaviour. Asking someone to describe a specific time they did something — not what they would do — forces them to retrieve a real memory with concrete details.
Examples
- "Tell me about a time you disagreed with your manager's decision. What did you do, and what was the outcome?"
- "Describe a project that failed. What happened, what was your role, and what did you learn?"
- "Give me an example of a time you had to learn something quickly under pressure."
What to look for: Specific details (names, dates, numbers). Vague or hypothetical answers ("usually I would...") signal they may be fabricating.
High predictive validity
Work sample / situational questions
Give candidates a scenario directly related to the job and ask how they'd handle it. Unlike behavioural questions, these test how someone thinks about your specific context — useful for roles where direct experience is rare.
Examples
- [For sales] "A prospect tells you your price is 40% higher than a competitor. Walk me through your response."
- [For engineering] "You're asked to fix a bug in production that's affecting 10% of users. No one knows the codebase. What's your first hour?"
- [For product] "DAU has dropped 15% for three consecutive weeks. Where do you start?"
What to look for: Structured thinking, not the "right" answer. How do they break down the problem? What questions do they ask?
Moderate predictive validity
Motivation and learning questions
These aren't about skills — they're about whether the person will stay and grow. Retention and performance over 12+ months correlates with intrinsic motivation, not just skill level.
Examples
- "What's the most technically complex thing you've learned in the last 12 months? How did you learn it?"
- "What kind of work do you find genuinely hard to stop doing?"
- "Tell me about a project where you went significantly beyond what was asked. Why?"
What to look for: Energy and specificity. Great candidates light up when talking about work they care about.
Questions with low or negative predictive value
"Where do you see yourself in 5 years?"
Candidates who answer well are those who've memorised the right answer. Nobody knows where they'll be in 5 years. This tests compliance, not performance.
"What's your greatest weakness?"
Universally answered with disguised strengths ("I work too hard," "I'm a perfectionist"). Tells you nothing about actual performance except whether they've heard this question before.
"Tell me about yourself."
Assesses rehearsal, not capability. If you want a summary, read their resume. Use this time to ask something real.
"Why do you want to work here?"
Tests whether they've read your about page, not whether they're competent. Save for culture-fit rounds, not first screens.
How to structure a 15-minute first round
A first-round screen doesn't need to be exhaustive. It needs to answer one question: Does this person deserve more of our time?
0–2 min
Context setting
Brief intro, set expectations. Not an evaluation stage.
2–6 min
1 behavioural question
Pick the most role-relevant trait. Listen for specificity.
6–10 min
1 situational question
Something directly tied to your actual work context.
10–13 min
1 motivation question
What do they genuinely care about? Does it match the role?
13–15 min
Candidate questions
What they ask tells you a lot too.
Why AI interviews produce better data
The core advantage of a structured AI first round is not speed — it's consistency. Every candidate gets the same questions in the same order. There are no interviewers having a bad day, no casual chit-chat that creates false affinity, no variation in how follow-up questions are asked.
This consistency is what makes the data actionable. When you're comparing 40 candidates all assessed by the same instrument, you can actually rank them. When each was interviewed by a different recruiter asking different questions, your shortlist is noise.
The takeaway
Stop asking what candidates would do. Ask what they've actually done. Stop testing interview preparation. Test job-relevant thinking. And make it the same for every candidate — that's when your first round starts generating signal.