"I just get a feel for people": why structured interviews win

A sales head interviewed nine candidates for two roles and picked the two he described as obvious. Both left inside eight months. Asked afterwards what had stood out about them, he said they were sharp, they got it, and he had known within ten minutes. He had thirty years of experience and a genuine instinct for people, and he had just made two expensive mistakes in a row.

The uncomfortable finding, consistent across decades of research and every business that has bothered to look at its own data, is that confidence in an interview judgement has almost no relationship to its accuracy. Experienced interviewers are more confident. They are not reliably more correct.

What an unstructured interview actually measures

An interview with no fixed questions, where the conversation goes wherever it goes, measures something real. It is just not the thing anybody intends to measure.

• Rapport. Whether the candidate is easy for this particular interviewer to talk to. Useful for a handful of roles, irrelevant for most, and heavily shaped by shared background, language and manner.
• Articulacy under low stakes. Fluent explanation of past work is a skill, but it is a different skill from doing the work, and the gap between them is where most bad hires live.
• Similarity. The strongest and least visible effect. Interviewers rate people like themselves higher, and then produce entirely sincere reasons that have nothing to do with the similarity.
• The first four minutes. A judgement forms early and the rest of the conversation is spent gathering support for it, which is why the interview can feel so confirming.

None of this makes interviewers bad people. It makes them people. The point of structure is not to distrust the interviewer; it is to give a normal human judgement something reliable to work on.

The whole intervention, in four parts

1  SAME QUESTIONS, SAME ORDER, EVERY CANDIDATE
     Four to six questions tied to the role. Follow-ups
     can vary; the core questions do not.

2  AGREED RATING SCALE, WRITTEN BEFORE YOU START
     What a 1 sounds like, what a 3 sounds like, what a 5
     sounds like - in words, for each question.

3  SCORE INDEPENDENTLY, THEN DISCUSS
     Each interviewer writes their scores before anyone
     speaks. Non-negotiable.

4  EVIDENCE, NOT IMPRESSION
     A score has a note attached saying what the candidate
     said that produced it.

That is the entire method. It is not sophisticated, it takes about two hours to set up for a role — which is the practical core of what interviewing skills for managers training covers, and it reliably outperforms the instinct of someone who has been hiring for three decades. The resistance it meets is almost never about efficacy — it is about the implication that experience is not enough.

Why independent scoring matters more than it sounds

Of the four parts, the third is the one businesses skip and the one that carries the most weight.

In an ordinary debrief, the most senior person speaks first. Everyone else's view then adjusts, mostly unconsciously, and the panel arrives at a consensus that feels robust and is in fact one person's opinion with three signatures on it. If the senior person is right, the panel added nothing; if they are wrong, the panel added nothing and removed the check.

Writing scores down before the discussion costs ninety seconds and preserves the disagreement. Disagreement is the useful part — two interviewers who scored the same answer 2 and 5 have heard different things, and finding out why is where the real information is.

Questions that work

The best predictor available in an interview is a specific account of something the candidate has actually done, probed for detail. Hypotheticals — what would you do if — measure how well someone reasons about a situation they are not in, which is a weaker signal and easier to rehearse.

1. 1

   Ask for a specific instance

   "Tell me about a time a delivery was going to miss its date. What did you do?" Then insist on one occasion rather than a general description of approach. "Usually I would..." is the answer to a different question.

2. 2

   Probe for their part in it

   "We" is the most common evasion in interviews, often unintentional. "What did you personally do?" and "what would have happened if you had not been there?" separate the contributor from the bystander without hostility.

3. 3

   Ask what went wrong

   Every real account has a difficulty. A candidate who describes only smooth successes is either describing someone else's work or has not reflected on their own.

4. 4

   Ask what they would do differently

   The single most informative question in most interviews. It reveals whether the person learns, and it is very hard to answer convincingly without having genuinely done the thing.

Scoring the answer, not the person

A scorecard only helps if the scale means something. "Rate communication 1 to 5" produces five interviewers using five private definitions, and the numbers that result are less useful than a conversation would have been.

Writing the anchors is the work. For a question about handling a missed deadline, a 5 might be: named a specific instance, described what they personally decided, said who they told and when, and identified what they would change. A 2 might be: described a general approach with no instance, or described the situation without their own role in it. Written like that, two interviewers who have never met tend to score the same answer within a point of each other — which is the entire objective.

Keep the notes short and factual: what the candidate said, not what you concluded about them. "Said he escalated to the client on day two" is evidence. "Seemed proactive" is the impression the evidence was supposed to support, and it is what makes debriefs unresolvable when two people disagree.

What to leave out

A structured process also makes it easier to notice questions that should not be asked at all — about marriage or family plans, about religion or caste, about age where it is not a genuine requirement of the role. These come up in unstructured interviews as small talk, usually without ill intent, and they create both an unfair process and an avoidable risk.

A fixed question set removes the opportunity almost entirely, which is a side benefit worth having. Where a business wants to be confident its hiring practice is defensible as well as effective, that is worth confirming with a qualified professional rather than settled from a blog post — but the structural fix is the same either way, and it also does most of the work that diversity, equity and inclusion commitments require in practice rather than in a policy document.

Making it survive contact with busy managers

The method fails in adoption rather than in design. Managers agree it is sensible, use it for the first two candidates, and then revert under time pressure — particularly the senior ones, who experience it as a constraint on judgement they have earned.

Three things help. Build the question set with them rather than handing it over, because a scorecard someone helped write is one they defend. Keep it to five questions, because a fifteen-question form will be abandoned. And show them their own data after two quarters — the correlation between interview scores and how the hire actually performed is the argument that lands, and it is the one nobody can make for them.

Beyond that it is a teachable skill like any other, and interviewing skills for managers is one of the few training investments with a directly traceable return, because the cost of a wrong hire is already known to the business. It is also worth checking that the rest of the funnel is not undoing the work: a strong interview process attached to a vague brief and an inconsistent shortlist still produces bad outcomes, which is where recruitment process consulting usually starts.