Structured Interviews Explained: Why They Beat Unstructured Ones

A structured interview takes away the freedoms most interviewers enjoy: picking questions on the fly, following tangents and settling on an overall impression. Giving them up is much of why it works. When Paul Sackett and colleagues re-analyzed decades of hiring research in 2022, structured interviews predicted job performance better than any other method they compared, and roughly twice as well as unstructured interviews.¹

Structured interviews ask all candidates identical questions, taken from an analysis of the job, and score each answer against standards written in advance. Because everyone is measured the same way, interviewers agree with each other more often, and their scores track job performance more closely than impressions from a free conversation do.²

In practice, that means four habits:

1. Write the questions from what the job requires.
2. Ask every candidate all of them, with any follow-ups planned in advance.
3. Rate each answer on a scale that describes weak and strong answers.
4. Have each interviewer record scores alone before any discussion.

Structured interviewing is step four of designing a hiring process.

What makes an interview structured?

An interview is structured when rules set in advance govern both what is asked and how answers are judged. The fullest framework, proposed by Michael Campion, David Palmer and James Campion in 1997 and reviewed by Julia Levashina and colleagues in 2014, lists 15 components in two groups: the content of the interview and the evaluation of answers.³

On the content side, questions come from an analysis of the job, every candidate hears the same ones, prompting and follow-up are limited, and candidates’ own questions wait until the end. On the evaluation side, each answer gets its own rating on a scale with written examples, interviewers take notes, several interviewers take part without comparing views between interviews, and the ratings are added up by a set rule instead of by gut feel. The components research uses most are job analysis, the same questions, better question types, anchored rating scales, a rating for every question and interviewer training.³

Take two managers hiring for the same support-team role. One asks whatever comes to mind and ends with a feeling. The other asks every candidate the same five questions about difficult customers and clashing deadlines, and circles a score after each answer. Only the second can set two candidates side by side and say where one was stronger.

What this means for you: structure is not one switch but a set of separate decisions, so you can add components one at a time, starting with those research uses most.

Why structured interviews beat unstructured ones

One big reason structured interviews predict better is that they measure more consistently. US federal hiring guidance from the Office of Personnel Management (OPM) says rules for asking, observing and scoring raise agreement between interviewers by limiting each one’s discretion, and that questions built on the skills critical to the job show high validity and agreement.²

A measure that disagrees with itself cannot predict anything well. In an open conversation, two interviewers meeting the same candidate ask, notice and weigh different things. One spends twenty minutes on travel stories while the other presses on budgets, so each verdict rests on different evidence.

Consistency may not be the whole story. A 2004 analysis by Frank Schmidt and Rodney Zimmerman found mixed support for the idea that it explains the entire gap. It also found that averaging three or four independent unstructured interviews matched the validity of one structured interview run by a single interviewer.⁴ Averaging smooths out noise after the fact; structure tries to prevent it.

In practice, treat wide disagreement between interviewers as a warning about the process, not as a healthy variety of opinion.

How much structure is enough?

Interview validity climbs steeply as questions and scoring become standardized, then levels off before the most rigid, word-for-word format. That was the central finding of a 1994 meta-analysismeta-analysis: A study that combines the results of earlier studies on the same question into one overall estimate. Pooling makes the estimate more precise, but it cannot repair the studies it pools: a meta-analysis of surveys is still survey evidence.Full entry in the glossary by Allen Huffcutt and Winfred Arthur, which sorted interviews by two separate dimensions: how standardized the questions were and how standardized the scoring was.⁵

Questions left openSame questions for allOne overall impressionEach answer scored against set standards

One overall impression, Questions left open: A conversationlowest structure

One overall impression, Same questions for all: Same questions, gut verdictcomparable input, loose judgment

Each answer scored against set standards, Questions left open: Free questions, careful scoringno studies found in the 1994 data

Each answer scored against set standards, Same questions for all: Structured interviewthe highest level in the 1994 study

The study combined the two dimensions into four levels, from a free conversation to a fixed script.

The study

Moderate evidence

Where extra structure stops paying: Huffcutt and Arthur, 1994

The authors pooled 114 validity results from interviews for entry-level jobs, each judged against supervisors’ ratings of job performance, and grouped the interviews into four levels of structure. Estimated validity, the correlationcorrelation: A measure of how closely two things move together, running from minus one, where one rises as the other falls, through zero, meaning no link, to plus one. It says how strong the relationship is, not what causes it, and it is not a percentage.Full entry in the glossary between interview scores and performance, rose from about 0.20 at the lowest level to about 0.56 at the third, then barely moved at the fourth, where every candidate got identical questions with no follow-up at all.⁵

The third level was already highly structured, typically with precise questions and scoring against set criteria, but it still allowed some variation, such as choosing questions or following up. The authors suggest one possible reason the extra rigidity added so little: a well-trained interviewer may learn more by probing a standard question carefully.⁵ The data covered entry-level jobs only, and the statistical corrections behind estimates from that era have since been argued to overstate validity.¹ A 2014 update by Huffcutt and colleagues, using a different correction method, still found the third and fourth levels about equal.⁶

Length is not the lever either: a 2018 meta-analysis by Thomas Thorsteinson found interview length unrelated to reliability or validity.⁷

1. A free conversation: open questions and one overall impression
2. Some structure: some formal rules, such as set topic areas or scoring on several criteria
3. High structure: usually precise questions and scoring against set criteria, with some room to choose or follow up
4. A fixed script: identical questions, no follow-ups, each answer scored against benchmark answers; little or no gain over step 3

So you do not need a script read word for word; the four habits listed near the start describe a high level of structure without one.

Past behavior or hypothetical: choosing the question type

OPM’s guidance describes situational questions, which put a hypothetical job situation to the candidate and ask how they would handle it, and behavioral description questions, which ask what the candidate actually did in a past situation. OPM says both have proven effective, and that past-behavior questions have shown higher validity where the work is highly complex, such as professional and managerial jobs.²

The two types seem to measure somewhat different things. Levashina’s review concluded that, when both are written to assess the same skill, situational questions mainly capture job knowledge or reasoning, while past-behavior questions mainly capture experience; pooled results show both predict performance, with past-behavior questions slightly ahead.³

For a team lead role, a past-behavior version might be: “Tell me about a time two urgent requests landed at once. What did you do, and what happened?” A situational version: “Two senior colleagues each need your help today, and there is time for only one. What would you do?” The review notes that researchers commonly recommend using both types, partly because they measure different things, and one study it cites found no sign that past-behavior questions lost validity with less job experience.³

Follow-up questions need rules too. The same review found that the research has not settled whether probing helps, and in one study probing increased faking, possibly because it signaled which answers mattered.³ So write the follow-ups in advance (“What did you do next?”, “What was the result?”) and use the same ones with everyone.

Scoring answers against a written scale

An anchored rating scale spells out what an answer at each point contains, from weak to strong, so the interviewer matches an answer to a description instead of to a feeling. In Levashina’s review, a meta-analysis of past-behavior interviews found higher validity and closer agreement between raters when anchored scales were used.³

Two further studies in the same review show what anchors do. With behaviorally anchored scales, job experts rated no better than non-experts. And a five-point scale with an example at every point made ratings less open to bias against applicants with disabilities than a scale anchored only at its ends, or not at all. The authors add that rating scales remain surprisingly under-researched.³

Anchors move the hard thinking about what good looks like to one calm session before interviewing, instead of repeating it after every answer.

The practical point: write the anchors before the first interview, from what good work in the role looks like, and test them on a colleague’s practice answers. If two raters put the same answer far apart, the anchors need sharper wording.

Several interviewers, scored separately

A panel, where several interviewers hear the same answers together, produces more agreement than separate interviews do, but not all of that agreement is accuracy. A 2013 meta-analysis by Allen Huffcutt, Satoris Culbertson and William Weyhrauch found average agreement considerably higher for panels than for separate interviews by different interviewers. The authors stress that such figures should count every source of measurement error, including how a candidate’s performance varies between occasions, which one panel sitting cannot show. They also flagged lower than expected agreement for highly structured interviews run separately.⁸

If a candidate has an off morning, every panelist sees it, the scores line up, and the agreement looks like reliability. The sterner test is whether separate interviewers on different days would agree.

OPM describes the usual way to use a panel: each member rates the answers individually, then the group discusses only to resolve large differences in scores.² Britain’s workplace advice service, Acas, suggests in its 2026 guidance that more than one person interview where possible, to reduce the risk of discrimination.⁹ Keep the format the same for every candidate too: a 2016 meta-analysis of twelve studies found that interviewers gave lower ratings in phone and video interviews than face to face.¹⁰

Structure and the law: why a consistent interview is easier to defend

A structured interview is easier to defend because it shows that candidates were treated alike and judged on the job. In a 1997 study of 130 discrimination claims decided in US federal courts, Laura Gollub Williamson and colleagues found that most aspects of interview structure were linked to verdicts for the employer, with objective, job-related content most strongly linked and standardized administration next. Their explanation: standardizing leaves less room to treat candidates differently, objectivity leaves less room for bias, and the interview is more clearly related to the job. The study is old and covers US federal courts only, so treat it as support for the logic, not a forecast of any case.¹¹

In practical terms, an employer who can produce the question list, the scored answers and the written anchors can show what was compared; a note saying a candidate “seemed a better fit” cannot.

The questions themselves also have to be lawful, and the rules depend on the country; the hiring-process guide lists what each protects. The US Equal Employment Opportunity Commission (EEOC) advises, as of September 2026, keeping pre-hire questions to what is essential for deciding whether a person is qualified, and says questions revealing race, sex, national origin, age or religion should generally be avoided, since they can be treated as evidence of an intent to discriminate.¹² Disability is stricter: the EEOC says an employer generally may not ask disability-related questions before making a conditional job offer, but may ask whether an applicant will need an accommodation to carry out a specific job duty.¹³

Section 60 of the Equality Act 2010, which applies in England, Scotland and Wales, forbids employers to ask applicants about their health before offering them work, with exceptions such as establishing whether the person can carry out a function intrinsic to the job; the Equality and Human Rights Commission alone enforces that ban, though acting on the answers can itself be disability discrimination.¹⁴ Acas’s 2026 guidance adds that the law requires employers to ask everyone invited to interview whether they need reasonable adjustments to attend, and calls it a good idea to put the same questions to each applicant where possible.⁹

US federal rules on selection procedures are also changing. In a July 2026 interim rule, OPM removed references to the 1978 Uniform Guidelines on Employee Selection Procedures from federal personnel regulations, citing a June 2026 Justice Department opinion that found them unlawful.¹⁵ Check current guidance and state or local law before relying on older summaries.

When a question needs HR or legal advice

Take a question to HR or an employment lawyer when the answer depends on legal detail you cannot confirm from official guidance. Now: a candidate objects to a question or says they were treated unfairly. Soon: before you add questions about health or adjustments, start recording interviews, or reuse a question set in another country. In Britain, Acas’s 2026 guidance says recording needs the candidate’s permission first, and notes and recordings must be stored securely under data protection rules.⁹ Routine: once a year, look again at the questions and anchors, and at who gets through the interview stage.

The bottom line

A structured interview works because it is consistent: the same job-based questions for everyone, answers scored against written anchors, and each interviewer’s scores recorded before anyone talks. You do not need a rigid script; most of the gain comes from those first steps. Keep the questions lawful where you hire, and ask HR or a lawyer when a decision turns on local rules.