Structured Interviews vs Unstructured Interviews: What's the Difference?

In 2013, three researchers ran an experiment that ought to have embarrassed the entire hiring profession. They asked people to interview strangers and predict their future academic performance, with one detail the interviewers were never told: some of those strangers had been instructed to answer at random, saying yes or no according to a hidden rule that bore no relation to the questions being asked.

‍

Almost none of the interviewers noticed anything amiss.

‍

They conducted what felt like real conversations, formed confident impressions of people who were in effect generating noise, and predicted their futures with the same easy assurance they brought to the genuine interviews. Worse, the interviews did nothing to sharpen anyone's predictions, and in some conditions blunted them; the people who skipped the conversation and worked from the records alone did at least as well.

‍

None of this would have surprised Paul Meehl. In 1954 he published a short book demonstrating that simple statistical formulas routinely matched or beat the judgement of trained expert clinicians, even when the experts had more information in front of them. The result has been reproduced across medicine, finance, parole decisions and university admissions for seventy years, and it always arrives at the same uncomfortable place: when a decision can be reduced to a few measurable factors combined by a consistent rule, the rule tends to win. Experts dislike this finding, understandably. So does anyone who has built a professional identity on the conviction that they can read other people.

‍

What structure actually changes

‍

The hiring interview is the most cherished version of that conviction. A structured interview is, in effect, an attempt to drag the conversation towards the statistical end of Meehl's spectrum: identical questions for every candidate, written in advance against the real demands of the role, and scored on a fixed scale so that two interviewers watching the same answer reach roughly the same number. An unstructured interview is the holistic alternative, the open conversation in which a perceptive human being supposedly synthesises dozens of fine signals into a single sound verdict.

‍

On the question of which approach better predicts how someone will actually perform in the job, the evidence is not close. Structured interviews are worth roughly twice as much, and that margin has held steady across decades of meta-analysis.

‍

The reason is the one the random-answer study laid bare. A free-ranging conversation throws off an enormous volume of information, almost none of it relevant to whether the person can do the work, and the human mind is more or less incapable of disregarding information once it has been presented with it. We fold everything into a story. The candidate's quickness, their handshake, the university on their CV, an unexpected shared loyalty to the same football club: none of it forecasts performance, and all of it gets absorbed into the impression and silently allowed to count.

‍

Understanding the role of human bias

This is where the least comfortable part of the argument lives. The information an unstructured interview smuggles in is random in its relevance, but not in its effects, because it consistently favours the familiar. Interviewers warm, without ever choosing to, towards candidates who share their background, their manner, their accent, the signals of a particular kind of upbringing, and that warmth presents itself as insight rather than affinity. A great deal of what selection panels describe as instinct, or chemistry, or “cultural fit,” is really the comfortable recognition of someone rather like themselves, dressed up in the language of judgement. The people who defend the human touch in hiring seldom mention that the human touch has always been the precise channel through which nepotism, in-group preference and other forms of discrimination travel. A fixed question and a fixed scale make nobody less prejudiced; they simply leave prejudice fewer places to hide.

‍

Why the gap is about to get expensive

‍

Beyond accuracy, there is a practical dividend. A process that asks the same things in the same order can be audited, compared and improved; you can look back over a year of hiring and see which questions did genuine predictive work and which were merely decorative. That is the real distinction between streamlining the recruitment process and simply making it quicker.

‍

The legal climate is shifting the same way. As the qualifying period for unfair dismissal in the United Kingdom shortens and the cost of removing a poor hire rises, a documented and defensible selection decision stops being good manners and starts being a form of insurance. A scoring sheet can be produced and explained in a tribunal; a sincere good feeling about a candidate cannot.

‍

Where assessment changes the picture

‍

The deeper lesson of Meehl's seventy-year-old result is not that interviews should be abolished, but that they should be demoted. The conversation is a poor instrument for forming a first judgement and a genuinely useful one for interrogating a judgement already reached on firmer ground. That inversion is the thinking behind a modern behavioural system. Rather than meeting a candidate and then hunting for evidence to justify the impression, you begin with a short, validated assessment that establishes where the person is actually strong and where the risks sit, and you spend the interview pressure-testing those specific findings instead of manufacturing new noise.

‍

Thrive is built on that inversion. Its science layer, The Faculty, built and overseen by occupational psychologists, reads each role on its own terms, selects the assessments that genuinely bear on it and weights them towards the behaviours that predict success in that particular job, rather than handing every employer the same fixed battery and calling it bespoke. What comes back is not a raw psychometric score awaiting a specialist's interpretation but a behavioural translation, a plain-language account of how the person is likely to work, carrying validity flags that say how far each part of the picture can be trusted. Understood properly, [recruitment assessment software] of this kind is not a substitute for human judgement, but rather the thing that finally gives human judgement something solid to stand on.

‍

What keeps the unstructured interview alive is not evidence, of which there is essentially none in its favour. The interview flatters a belief most of us are unwilling to surrender: that we are shrewd judges of character, and that half an hour across a table is enough to take the measure of a stranger. Seventy years of data say we are not, and the better part of a century of expensive hiring mistakes says so too. The organisations now pulling ahead are simply the ones prepared to find that a little humbling, and to let a tested instrument do the work their intuition was never much good at to begin with.