Key takeaway: A structured interview is one in which every candidate for a role is asked the same predetermined, job-related questions, in the same order, and each answer is scored against a written rubric before the interviewers discuss. Decades of research, most cited from Schmidt and Hunter's 1998 meta-analysis of selection methods, found structured interviews predict job performance substantially better than unstructured ones, with validity roughly on par with work-sample tests.
Unstructured interviews, where each interviewer improvises questions and forms an overall impression, feel more natural and are much worse at predicting who will do the job well. They also give the most weight to whatever the interviewer happened to notice first.
What makes an interview "structured"?
- Questions are fixed in advance and derived from the competencies the role needs.
- Every candidate gets the same questions, so answers can be compared.
- Answers are scored on a rubric (for example 1-4 with written anchors for each level) immediately after the interview, before any discussion with other interviewers.
- Interviewers are trained on the questions and the rubric.
- Scores are combined mechanically (summed or averaged) rather than by gut feel.
Structure is a spectrum. Even adopting the first three steps moves a process most of the way toward the predictive end.
What kinds of questions are used?
- Behavioural: "Tell me about a time you had to ship with incomplete requirements. What did you do and what happened?" Past behaviour as the predictor.
- Situational: "A customer escalates to your manager about a bug you deprioritised. What do you do?" Hypothetical judgment.
- Work sample or technical: A task that resembles the job, scored on a rubric.
Each question should map to one competency, and each competency should have a question. A scorecard lists the competencies, the questions, and the anchors.
How to run one in practice
- Write the scorecard: five to seven competencies, one or two questions each, and 1-4 anchors.
- Assign questions across interviewers so nothing is asked twice and nothing is missed. See interview panels.
- Each interviewer submits scores and evidence (what the candidate actually said) within a few hours, before the debrief.
- The debrief reviews scores and evidence, not impressions.
- Keep the scorecards; they are the record if a decision is ever questioned.
Where Noon fits
Noon's AI Interviewer runs structured first-round interviews from a scorecard and produces consistent, comparable write-ups, and its AI Notetaker captures and summarizes human interviews against the same competencies. Both come with Noon's unlimited plan (pricing). Noon does not make hiring decisions; it gives interviewers better records to make them.
Frequently asked questions
Are structured interviews more accurate than unstructured ones?
Yes. The 1998 Schmidt and Hunter meta-analysis estimated the predictive validity of structured interviews at about 0.51 versus about 0.38 for unstructured interviews, a difference that has held up in later reviews.
Do structured interviews feel robotic to candidates?
They do not have to. Interviewers can build rapport, ask follow-ups for clarification and let the conversation breathe; what stays fixed is the core question set and the scoring.
How many questions should a structured interview have?
Enough to cover five to seven competencies in the time available, usually five to eight core questions in a 45-60 minute interview, each with follow-ups.
Can structured interviews reduce bias?
They reduce the room for it. Fixed questions, written anchors and scoring before discussion all limit the influence of first impressions and similarity to the interviewer. They are not a complete solution on their own.