Structured vs unstructured interviews, measured again

On 16 July 2026 we published a post on why your phone screen can’t predict performance. It said unstructured phone screens predict job performance at r = .38 and structured ones at r = .51. Those figures were correct for the source we used, Schmidt and Hunter’s 1998 meta-analysis. They were also out of date. A 2022 re-analysis lowered both numbers, and it lowered the unstructured one twice as much. If anything, the case for structure got stronger.

This post is the correction, and a fuller answer to structured vs unstructured interview: what each one is, what the newer research says, and why the round most hiring teams leave unstructured is the very first one. It is part of a series on how one voice interview replaces the recruiter phone screen.

Structured vs unstructured interview: what is the difference?

A structured interview asks every candidate for the same job the same job-related questions and scores the answers on the same written rating scale. An unstructured interview lets the interviewer choose questions as the conversation goes and judge the candidate on an overall impression. Structured interviews predict job performance roughly twice as well.

That last sentence depends on the research, so here it is.

What the 2022 research found

In 2022, Paul Sackett, Charlene Zhang, Christopher Berry and Filip Lievens published a re-analysis of the validity of selection methods in the Journal of Applied Psychology. They argued that earlier studies had over-corrected for range restriction, which inflated many of the famous numbers. With the corrections redone, structured interviews came out at an operational validity of .42, the highest of the selection methods they examined, ahead of cognitive ability tests at .31 (summary by SIOP).

Unstructured interviews fell further. A comparison table from Master International sets the two studies side by side:

In plain words, a validity coefficient tells you how strongly interview scores line up with later job performance. Squaring it gives a rough share of the differences in performance the interview accounts for. That is about 18 percent for a structured interview at .42 and about 4 percent for an unstructured one at .19. So the corrected gap is not a small edge. An unstructured interview carries a fraction of the information.

Two cautions. These are averages over many studies, and a badly built structured interview can do much worse than .42. And validity is about prediction, not fairness or candidate experience, which need their own checks.

What makes an interview “structured”

Structure is not one thing. It is a set of choices, and most interviews have some of them and not others. Google’s guide to structured interviewing names the core of it as vetted, role-relevant questions, the same questions for every candidate, and a standard rubric so every reviewer shares one idea of what a good answer looks like.

A practical checklist:

1. Questions come from the job. Written from what the role actually does, before the first interview.
2. Every candidate gets the same core questions. Follow-ups can vary; the core cannot.
3. Each question has a written rating scale. Anchors describe what a 1, a 3 and a 5 sound like.
4. The interviewer scores each answer, not the person. One score per question, not a single verdict at the end.
5. Notes record what the candidate said. Evidence, not adjectives.
6. Scores are written down before anyone discusses the candidate. Otherwise the loudest opinion sets the score.

Google also reports that pre-made questions, guides and rubrics save an average of 40 minutes per interview. Structure is faster, not slower, once it is written.

Why the first phone screen is the least structured interview you run

Most hiring teams put real effort into structuring their final rounds. The first round, the recruiter phone screen, usually gets the least. There is no fixed question list, the recruiter adapts to whoever picks up, notes are a line or two, and the outcome is “seems good” or “not a fit.”

SHRM’s advice on effective phone screens says to rank the job’s key factors and use the same criteria with every candidate, and to avoid judging on likability. Those are the exact two things an improvised call struggles with.

This matters more than it looks. The first screen sees every candidate, so it is where the most people are rejected. The least predictive interview in the process makes the most decisions. A candidate who would have scored well in a structured final round never reaches it because of a 12-minute chat with a tired recruiter at 6 PM.

How to add structure without sounding like a robot

The usual objection is that structured interviews feel stiff. They do when structure is mistaken for reading a script word for word. Here is a version that keeps the conversation human:

• Fix three to five core questions per role. For a customer support role: one on handling an upset customer, one on a time they missed something, one on why this job.
• Allow one follow-up per question, chosen by the answer. “What did you say next?” or “What was the result?” Follow-ups dig into the same question; they do not open new topics.
• Write anchors for each question. For the upset-customer question: 1 = blames the customer or gives a generic answer; 3 = a real example with a sensible fix; 5 = a real example, a fix, and what they changed afterwards.
• Score immediately after each answer, not at the end of the call.
• Keep small talk outside the scoring. Rapport is fine. It is just not evidence.

A semi-structured interview built this way keeps most of the benefit. Google’s guide reports that rejected candidates who had a structured interview were 35 percent happier than those who did not. Being asked fair, relevant questions feels better than a chat that wanders.

Where an AI voice interview fits

Structuring the final round is a training problem. Structuring the first round for 800 applicants is a capacity problem, because each recruiter call takes 10-15 minutes and the structure slips as the day goes on.

HireQwik runs that first round as a 10-20 minute voice interview the candidate takes in a browser at a time they book. It is structured by construction. The questions come from the job, plus the candidate’s own resume, under one setup per job, so the 1st and the 400th candidate are measured against the same job on the same scale. A conversational mode asks up to 12 questions naturally and follows up when an answer is vague. Scoring happens after the call from the full transcript, against a written 0-10 scale: 3-4 for vague answers with no specifics, 6-7 for specific examples with real detail, and 8 or more only for genuinely standout answers. Every score quotes the candidate’s own words, and the scoring is set for low randomness, so the same answer is scored consistently.

The result is a recommendation (Strong Go, Go, On Hold or No Go), not a decision. HR decides who moves forward, and can read or listen to any interview before it does. The final round stays human, and now it starts from a structured first round. To hear what the candidate hears, try the live AI interview. If you are building scoring sheets for the later rounds, our interview scorecard page has a template.