Structured vs unstructured interviews: what the evidence actually says

Most hiring decisions still come down to a conversation. Someone walks in, you talk for half an hour, and you leave with a feeling about whether they can do the job. That feeling is confident, it is specific, and decades of research say it is mostly wrong.

The interview itself is not the problem. The way most interviews are run is. When researchers compare a free flowing conversation with a structured interview, where every candidate answers the same questions and is scored against the same rubric, the structured version predicts job performance more than twice as well [1]. This piece walks through what that evidence shows, why the casual interview fools even experienced interviewers, and what a defensible interview looks like in practice.

What counts as a structured interview

A structured interview is not a stiff interview or a longer one. It is an interview built on four design choices that remove guesswork and make candidates comparable.

1. The same questions for everyone. Every candidate for a role is asked the same job related questions in the same order, so you are comparing answers, not improvising a different conversation with each person.
2. A rubric defined in advance. Before anyone is interviewed, you decide what a strong, average, and weak answer looks like for each question.
3. Independent scoring against that rubric. Each answer is rated on the rubric as it happens, not blended into one gut impression at the end.
4. Evidence behind every score. The rating points back to what the candidate actually said, so the decision can be reviewed later.

An unstructured interview drops all four. Questions vary by candidate and interviewer, there is no rubric, and the outcome is a holistic judgment formed in the moment. That single difference, structure or no structure, is one of the most studied variables in all of personnel selection, and the gap it produces is large.

What the validity numbers actually show

In selection research, a method is measured by its operational validity: how strongly it predicts who will actually perform on the job, on a scale where 0 is no better than chance and 1 is perfect prediction. A landmark 2022 reassessment of a century of hiring research put the structured interview at the top of the practical methods [1].

At 0.42 against 0.19, a structured interview predicts performance more than twice as well as the same conversation run without structure [1]. It also edges out the cognitive ability test, which for years was treated as the gold standard. The striking part is that both interviews involve the same people, the same time, and the same room. The only thing that changes is the discipline of the process.

A separate line of research explains why structure helps. Across more than 150 studies, when you combine candidate information mechanically, with a consistent rule or rubric, rather than by holistic judgment, prediction of performance improves by more than 50 percent [2]. Experts adding their intuition on top of the rubric did not help. They made it worse.

Why the unstructured interview feels better than it is

If unstructured interviews predict so little, why do they feel so revealing? Because the human brain is very good at manufacturing a confident story from almost nothing.

In one striking study, interviewers questioned candidates who had been secretly instructed to answer some questions completely at random. The interviewers still formed coherent, confident impressions of those candidates, and almost none of them noticed anything was wrong [3]. The conversation produced a feeling of insight that the answers did not justify.

It gets earlier than the answers, too. In real hiring interviews, the impression a candidate makes in the first few minutes of small talk, before any substantive question is asked, predicts the final interview rating and even who gets the offer [4]. In an unstructured interview there is nothing to catch this. The rapport becomes the result.

The casual chat feels insightful and predicts almost nothing. Structure is what turns an interview from a first impression into evidence.

The four ingredients, and what each one fixes

Each design choice in a structured interview targets a specific way the unstructured version goes wrong.

Same questions, asked of everyone

Identical questions make answers comparable and stop the interview from drifting toward whatever a confident candidate steers it to. A quiet applicant and a polished one get measured on the same things.

A rubric agreed before anyone is interviewed

Deciding what good looks like in advance is what removes the moving goalposts. Without it, the bar quietly shifts to flatter whoever is in the room, which is exactly how bias slips in unnoticed [6].

Scoring each answer on its own

Rating answer by answer, rather than forming one overall feeling at the end, is the mechanical combination that the research credits with the biggest accuracy gain [2]. It stops one charismatic moment from coloring the whole evaluation.

Evidence behind every score

A score that points back to what the candidate said is one you can review, explain, and defend. It also makes the whole process auditable, which matters more every year as hiring regulation tightens.

What this means for high volume hiring

Structure is hardest to maintain exactly when it matters most. When one role draws hundreds or thousands of applicants, the temptation is to run quick, informal screens to get through the pile. That is the unstructured interview at scale, and it inherits all of its weaknesses while adding a fairness problem: different applicants get measured by different standards on different days.

This is the case for giving every applicant the same structured interview and scoring each one against the same rubric, with the evidence kept on record. That is exactly what Sage, our AI interviewer, is built to do: it asks your questions, scores each answer against your rubric with quoted evidence, and hands you a ranked shortlist, so structure survives the volume instead of collapsing under it. The same discipline applies before the interview too, to screening every resume against a rubric rather than skimming each one for a few seconds.

Putting it into practice

You do not need to rebuild your whole process at once. A few changes capture most of the gain.

• Write a short list of job related questions and ask every candidate the same ones.
• For each question, define in advance what a strong, average, and weak answer looks like.
• Score each answer against that rubric as the interview happens, and note the evidence.
• Combine the scores by a fixed rule rather than an overall gut feeling.
• Keep the record, so any decision can be reviewed and explained later.

None of this makes interviews robotic. It makes them fair and comparable, and it replaces a confident feeling with evidence you can stand behind. That is the whole difference between an interview that predicts performance and one that only feels like it does.

Muhammad Shahbaz Manzoor

AirbaseHQ

Muhammad writes about hiring, the evidence behind better decisions, and building AirbaseHQ.

LinkedIn

This article is for general information only and is not legal, financial, or professional advice. Laws and regulations vary by location and change over time, and statistics and research are drawn from third-party sources that may be updated or revised. For decisions that affect your organization, check the specifics with a qualified professional.