Structured interviews, a practical guide

What makes an interview structured?

An interview is structured when every candidate is asked the same job-related questions and every answer is rated against the same written standard. The benefit depends on doing both consistently.

A widely used framework, from Campion, Palmer and Campion (1997), lists 15 components of structure. A later review by Levashina, Hartwell, Morgeson and Campion (2014) groups them by whether they shape the content of the interview or the way answers are evaluated:

You don't need all 15. In the studies Levashina and colleagues reviewed, the components used most often were job analysis, the same questions, better question types, anchored rating scales, rating each question and training.

Why do structured interviews predict job performance better?

Because they remove much of the variation in what gets asked and how answers are judged, so the ratings say more about the candidate and less about the interviewer.

Researchers measure this as validity: the correlation between interview ratings and later ratings of job performance, corrected for known measurement problems. In a widely cited summary, Schmidt and Hunter (1998) reported a validity of .51 for structured interviews and .38 for unstructured ones.

In 2022, Sackett, Zhang, Berry and Lievens re-analyzed the evidence. They argued that an adjustment called the correction for range restriction, which accounts for the fact that validation studies mostly include people who were hired, had often been applied in ways that overstated validity. Their revised estimates were lower for most methods, and the order changed:

Structured interviews came out on top, and the gap between structured and unstructured interviews widened. The authors also stressed variation: the estimate for structured interviews had a standard deviation of .19 across studies, and a lower 80 percent credibility value of .18 (Sackett et al., 2023). An average of .42 hides interviews that work far better and far worse. Structure is not a guarantee, and a badly built structured interview can still predict poorly.

Structure also appears to help with fairness. In Levashina and colleagues' meta-analysis of studies published after 1996, differences in structured interview ratings by race, gender and disability were mostly negligible, though the authors cautioned that the number of studies was small.

How do you build a rubric with anchored ratings?

Write the competencies first, then the questions, then a description of what each rating sounds like for each question. The order matters:

1. Name three to five competencies the job needs, taken from the hiring brief or a short job analysis.
2. Write one or two questions per competency. Past-behavior questions ask what someone did ("Tell me about a time you…"); situational questions ask what they would do. Researchers commonly recommend using both (Levashina et al., 2014).
3. Anchor every point on the scale. For each question, describe what a weak, mixed, good and strong answer contains, using the kinds of things candidates actually say.
4. Plan your follow-ups. Decide in advance which neutral probes you will use, and use them the same way with everyone.
5. Try the questions on a colleague. Fix questions that confuse and anchors that don't separate answers.
6. Train the interviewers on the questions, the anchors and how to take notes.

Here is an anchored scale for one question: "Tell me about a time new information forced you to change a plan you'd committed to."

Anchors make a measurable difference. In one meta-analysis of past-behavior interviews summarized by Levashina and colleagues, interviews with anchored rating scales had higher validity (.35 against .26) and higher agreement between raters (.77 against .73). Another study they review found that a five-point scale with an anchor at every point made ratings more resistant to disability bias than a scale anchored only at its ends.

Why should interviewers score before the debrief?

Because once opinions are shared, the first and most confident voices can shape everyone else's memory of the interview. Independent scores preserve what each interviewer actually saw.

The US Office of Personnel Management's guide to structured interviews describes the sequence: each panel member reviews their notes and rates the candidate right after the interview, forming an independent evaluation "without discussion with other panel members", and only then do they compare notes and reach a consensus (US Office of Personnel Management, 2008). Campion and colleagues list no discussion of candidates between interviews as a component of structure in its own right.

A simple debrief follows from this:

1. Every interviewer submits scores and notes before the meeting.
2. Scores are revealed at the same time.
3. The panel discusses the largest disagreements first, using the notes.
4. The hiring manager records the decision and the reasons for it.

Rating as you go may also help. An exploratory analysis by Levashina and colleagues suggests that rating after each question, rather than once at the end, might reduce differences between groups.

What are the most common mistakes?

• Questions that drift. If the questions change from candidate to candidate, the ratings can't be compared.
• One overall impression. Scoring the whole interview at the end invites the halo of a good first five minutes.
• Anchors at the ends only. "Strong" and "weak" leave the middle of the scale to taste.
• Unplanned probing. Follow-ups that differ by candidate undo the structure. Levashina and colleagues recommend planned, neutral probes, used equally, and only when an answer is clearly incomplete or off track or the candidate is shy.
• Leaking other information. Resumes, test scores and other interviewers' opinions can tilt ratings. Keep them out of the room until scoring is done.
• Talking between interviews. A hallway comment becomes a frame for the next interviewer.
• No training. A good rubric in the hands of untrained interviewers is still an unstructured interview.

Where does AI help, and where doesn't it?

AI helps by keeping the structure in place: asking the same questions, capturing what was said, and filing evidence under the right competency. Ratings and decisions should stay with trained people.

In Emoti, interview notes follow a template with one section per topic, and every point links back to its moment in the transcript, which makes it easier to score from evidence rather than memory. Emoti's first-round voice screens ask every candidate the same questions and rate each answer against a rubric the team writes, and a recruiter makes every decision. Both have limits that we describe openly in Grounded interview notes and Designing first-round voice screens candidates can trust.