Structured Interviews: A Practical Guide and Question Templates

Why unstructured interviews predict so poorly

In an unstructured conversation, the interviewer is the instrument: follow-ups, pet questions, the CV on the table, the small talk beforehand. Those degrees of freedom cut predictive validity. Schmidt and Hunter summarised 85 years of selection research and placed structured interviews at an operational validity of .51 — level with general mental ability tests, well above unstructured interviews (.38) [1]. McDaniel et al. (1994) showed in a meta-analysis of employment interviews that structure is the moderator that matters, not interviewer charisma [2].

• Same candidates, different interviewers: without structure you get rater drift, not “good intuition”
• Reading the CV during the interview pulls attention to pedigree instead of observed behaviour
• Clinical overall impressions (“culture fit”) are hard to defend if a discrimination claim follows

Read the numbers honestly: 1998 vs 2022

Many articles still quote “validity .51”. Sackett, Zhang, Berry, and Lievens (2022) re-examined the range-restriction corrections behind those figures and produced lower, more credible estimates [4]. The ranking holds: structured interviews remain among the strongest predictors. The absolute values are smaller — and that is the version you should use internally.

What structure is: Campion’s 15 elements

Campion, Palmer, and Campion (1997) decomposed interview research into 15 structure components — from job analysis through to how ratings are combined [3]. You do not need all 15 on day one. The eight below capture most of the validity gain and can live in an ATS.

• Derive questions from a job analysis — not from the first candidate’s CV
• Ask every candidate the same core questions in the same order
• Limit follow-ups to clarification, not new topics
• Use behavioural and situational questions instead of “tell me about yourself”
• Score each answer, rather than one overall impression at the end
• Put behavioural anchors on the scale (what is a 2 vs a 5?)
• Use multiple interviewers, score independently, then calibrate
• Combine ratings statistically (average) instead of “talking it out” in the corridor

STAR questions: a template for four core competencies

Behavioural questions treat past behaviour as the predictor (Situation – Task – Action – Result). Write two questions per competency so one weak example does not sink the whole dimension. Adapt these to your role:

• Collaboration: “Describe a time you disagreed with a stakeholder. What did you actually do — and how did it end?”
• Problem-solving: “Tell me about a piece of work that did not go to plan. Which options did you consider, and which decision did you take?”
• Ownership: “When did you last pick up something that was not formally yours? What triggered it, and what was the outcome?”
• Communication: “Walk me through how you delivered a difficult message to a non-expert audience. How did you know it landed?”
• Probe only along STAR: “What exactly did you say?”, “What was everyone else’s role?”, “How did you measure success?”

A scoring rubric teams will actually use

Without anchors, a 1–5 scale collapses into “everyone is a 4”. The SIOP Principles require selection procedures to be tied to job requirements and for judgment rules to be documented [5]. A rubric you can explain in 30 seconds:

How to introduce this in one week

Structure rarely fails from lack of knowledge. It fails because the first role looks “too special”. Start with one role, not a change programme.

• Day 1: 4–6 competencies from the role profile (not 12)
• Day 2: calibrate the question set and anchors with two hiring managers (15–20 min, one dummy example)
• Day 3: put the guide in the ATS so every interviewer sees the same form
• From interview 1: score independently, then 20 minutes of calibration — discuss gaps of ≥2 points
• After 5 candidates: drop questions that do not discriminate

Conclusion

Structured interviews are not an HR ritual. On current revised estimates they are the strongest selection procedure in the classic toolkit [4]. The work sits before the first conversation: questions, anchors, the same process. After that, every round gets faster and more defensible — both for the hire and if you later have to explain it.