Structured interviews: what they are and why they work

What makes an interview structured

Schmidt and Hunter's well-known 1998 review draws the line clearly. Unstructured interviews have no fixed format or set of questions, and the same interviewer often asks different people different things. Answers usually aren't scored one by one. The interviewer gives an overall rating based on an impression.

Structured interviews are the opposite. The questions are fixed, and they usually come from a careful look at the job. There is a set way to score each answer.

Structure isn't all or nothing. A 1997 review by Campion, Palmer and Campion identified fifteen components of structure that can improve either the content of an interview or how it's evaluated. You don't need all of them to get started. Three do most of the visible work:

• Questions based on the job, written before you meet anyone.
• The same questions, in the same words, for everyone.
• A scoring rubric agreed in advance and applied to each answer.

What the research says

Researchers measure this with validity: the correlation between a selection method and later job performance. Zero means no relationship, and one would be a perfect prediction.

In 1998, Schmidt and Hunter put structured interviews at .51 and unstructured interviews at .38. In 2022, Sackett, Zhang, Berry and Lievens argued that earlier work had overcorrected for a statistical problem called range restriction. Their revised estimates were lower for most methods: .42 for structured interviews and .19 for unstructured ones.

The revised gap between the two is wider than the old one. Structured interviews also came out with the highest mean validity of any method in that analysis, ahead of job knowledge tests at .40, work samples at .33 and general mental ability tests at .31.

What the research doesn't say

The .42 figure is an average, and the spread around it is wide. The same authors report an 80% credibility interval from .18 to .66, and they suggest reading the result as .42 plus or minus .24. A badly built structured interview can do much worse than a well-built one.

The studies also looked at interviews as a method, mostly run by people. Those numbers don't transfer automatically to any particular tool or AI product, including ours.

The authors reported group differences too. The Black-White mean difference was .23 for structured interviews, against .32 for unstructured interviews and .79 for general mental ability tests. Smaller isn't the same as none, so keep watching your own outcomes.

Why structure helps

The practical logic is easy to see. If two candidates answer different questions, you are comparing two conversations. If they answer the same question and you score it the same way, you are comparing the candidates.

Writing the criteria first also changes what the interviewer pays attention to. You decide what matters for the job before you have met someone you like.

How to build one

You can do a first version for one role in an afternoon.

• List what someone must do well in the first six months of the job.
• Pick the three or four things that matter most and can be judged in conversation.
• Write one or two questions for each, about real past work.
• Write down what a weak, adequate and strong answer sounds like.
• Try the questions on a colleague and fix anything that confuses them.
• Ask everyone the same questions and score each answer before moving on.
• Compare scores after the interviews, not impressions.

Common mistakes

Most structured interviews fail in small ways, usually under time pressure.

• Reading the script but giving one overall score at the end.
• Letting a good first answer colour the scores for the rest.
• Adding a few extra questions for candidates you already like.
• Using criteria nobody can define, such as culture fit.
• Keeping old questions after the job has changed.

The trade-offs

Structure has real costs. Schmidt and Hunter note that structured interviews cost more to build and use. They also feel stiffer. Candidates may find a script less warm, and an interviewer can't always follow a surprising thread.

A common answer is to use structure where comparison matters most, in the early rounds, and leave more room for open conversation later. People can then explore what the first round surfaced, and candidates get a chance to ask their own questions.

Where Vettika fits

Vettika runs the first round as a structured interview. The AI interviewer reads your questions, so every candidate gets the same words, and AI scores the transcript against your rubric, criterion by criterion. It asks no follow-up questions, which keeps the comparison fair but means your team does the probing in the next round.

Sources

• Schmidt and Hunter (1998), The validity and utility of selection methods in personnel psychology, Psychological Bulletin
• Sackett, Zhang, Berry and Lievens (2022), Revisiting meta-analytic estimates of validity in personnel selection, Journal of Applied Psychology
• Sackett, Zhang, Berry and Lievens (2023), Revisiting the design of selection systems, Industrial and Organizational Psychology
• Campion, Palmer and Campion (1997), A review of structure in the selection interview, Personnel Psychology

Common questions