Ask most hiring managers whether they trust their read on a candidate after a good conversation, and they'll say yes, confidently. Ask personnel-selection researchers whether that confidence is justified, and the answer, replicated across decades of studies, is no — unstructured interviews are one of the weaker predictors of job performance available, while structured interviews are among the stronger ones, and the gap between them is large enough to matter for every hiring decision made without structure.
What "structured" actually means, precisely
A structured interview asks every candidate for a given role the same set of predetermined questions, in the same order, and scores each answer against a rubric defined before any candidate was interviewed — not against a general impression formed afterward. This is a narrower and more specific definition than "having a list of topics to cover." A structured interview with a vague rubric, or one where interviewers feel free to skip questions that "don't seem relevant" for a particular candidate, has already lost most of the structure that produces the predictive-validity gain.
The mechanism: removing inconsistency and post-hoc rationalization, not asking better questions
The predictive power gain doesn't come from structured interviews asking magically better questions than unstructured ones. It comes from removing two specific sources of error: interviewer-to-interviewer inconsistency (different interviewers unconsciously weighting different things, making the same candidate look strong to one person and weak to another), and post-hoc rationalization (forming a gut impression in the first few minutes, then unconsciously interpreting the rest of the conversation to confirm it — a well-documented cognitive bias in interview settings). A shared question set and a rubric defined in advance directly targets both of these, independent of whether any individual question is cleverly designed.
The rubric has to exist before the interview, not after
The single most common way structure gets undermined in practice: interviewers keep a consistent question list but score candidates by overall impression after the conversation, rather than against criteria written down beforehand. This reintroduces exactly the post-hoc rationalization problem structure is meant to solve — a rubric written after hearing the answers is shaped by the impression already formed, not an independent yardstick. A rubric has to specify, in advance, what a strong, adequate, and weak answer to each question actually looks like, ideally with concrete example responses at each level.
Structure doesn't mean sounding like a script
A common objection: structured interviews feel robotic and hurt candidate experience. This conflates the wording of questions with the fact of asking the same ones. The delivery can be entirely conversational — an interviewer can phrase the same underlying question naturally, follow up genuinely, and build real rapport — as long as every candidate for the role is asked the same core questions and scored against the same predefined rubric. What has to stay fixed is the substance and the scoring, not the tone.
What a structured process actually requires
- A fixed set of questions per role, ideally tied to specific, job-relevant competencies rather than generic ones
- A written scoring rubric with example answers at each score level, finalized before any candidate is interviewed
- Every candidate for a given role gets the same questions, in the same order, with no interviewer discretion to skip ones that "don't seem relevant"
- Scores recorded immediately after each answer, not reconstructed from memory at the end of the conversation
None of this eliminates interviewer judgment — it constrains where that judgment gets applied, from "did I like this person overall" to "how well did this specific answer meet this specific, predefined bar." That constraint is the entire mechanism, and it's a process change, not a smarter-questions change.