Structured vs unstructured interviews: what the evidence actually says.

Structure correlates higher.

Structured interviews correlate with job performance at about .42; the unstructured interview at about .19 (Sackett et al., 2022). Work samples sit at .33. These are correlations, not hit rates. They are corrected estimates pooled across decades of studies, not a head-to-head trial. The gap is large and consistent. The exact figures carry error bars.

Source: Sackett et al., 2022 — revised meta-analytic estimates →

Why intuition loses.

Interviewers form a coherent story about a candidate within the first few minutes and spend the rest of the conversation confirming it. In Dana, Dawes and Peterson (2013), the people being interviewed secretly answered yes-or-no questions at random, using the first letters of the last two words of each question to decide. The student interviewers never noticed, and rated those interviews as informative as the real ones. The feeling of insight arrives whether or not there is anything to see.

A structured interview, or a scored work sample, removes the room where that story gets built. Every candidate gets the same job-relevant task; every answer gets the same rubric; the score is recorded before candidates are compared.

Autopilot isn’t the fix either.

Handing hiring to an off-the-shelf AI screener is not the structured version. Wilson & Caliskan (AIES 2024) audited three Mistral-based open models on résumé screening and found white-associated names preferred in 85.1% of comparisons. Structure without a human on the call automates the bias.

Source: Wilson & Caliskan, AIES 2024 →

Score by machine, verify by human.

An algorithm scores the work sample consistently (Kuncel 2013 shows mechanical combination beats holistic judgment); a recruiter verifies authenticity, handles exceptions, watches adverse impact, and owns the call. Hoffman (2018) showed overriding a valid test hires worse. Vaccaro (2024) showed averaging human and AI often loses. So divide the labour rather than averaging it. BitStage sells software that does this, which is a reason to check the sources above rather than take the recommendation on trust. The evidence rules out the manager alone and rules out blending the scores. It does not establish that this particular split is optimal.

What is a structured interview?

A structured interview asks every candidate the same job-relevant questions in the same order, scored against a defined rubric before candidates are compared. The interviewer's job is to observe and score, not to synthesize a gut feel.

What is an unstructured interview?

An unstructured interview is a conversation. Questions vary by candidate, scoring is holistic, and the decision is a synthesis of impressions.

Which is more predictive of job performance?

Structured interviews correlate with job performance at about .42; the unstructured interview at about .19 (Sackett et al., 2022). Work samples sit at .33. These are correlations, not hit rates.

Why does intuition fail in unstructured interviews?

Interviewers form impressions in the first minutes and spend the rest of the conversation confirming them. Dana, Dawes, and Peterson (2013) showed that interviewers form confident, coherent narratives even from random answers — the feeling of insight is not evidence of it.

Are structured interviews biased?

All hiring methods carry bias risk. Structured formats reduce interviewer-specific bias by fixing the questions and rubric, and they make disparate impact measurable — which is the pre-condition for fixing it. Unstructured formats hide the bias inside a gut feel.

What does BitStage recommend?

BitStage sells this, so weigh it accordingly. Replace the frontline phone screen with a short, job-relevant role rehearsal, scored consistently by an algorithm and verified by a recruiter who owns the call. The evidence rules out trusting the manager alone and rules out blending the scores; it does not establish that this particular split is optimal.