The free-flowing interview is the most trusted tool in hiring and one of the least reliable. Decades of research point the same way: unstructured interviews predict job performance poorly, largely because interviewers form an impression in the first minutes and spend the rest of the conversation confirming it.

Structure fixes most of that, and it does not require making the conversation mechanical.

Define what you are assessing before you meet anyone

Start from the job, not the CV. List the four or five capabilities that genuinely determine success — for a payroll officer that might be accuracy under deadline, statutory knowledge, discretion with confidential data, and the ability to explain a payslip to an upset employee.

Those become your assessment criteria. Everything in the interview exists to gather evidence on them, and anything that does not is conversation rather than assessment.

Ask every candidate the same core questions

Same questions, same order, for everyone. This is what makes comparison possible — otherwise you are comparing impressions formed from different conversations, which is not comparison at all.

Follow-up probes can and should differ. It is the core set that stays fixed.

Ask about what happened, not what they would do

Hypothetical questions test imagination. Past-behaviour questions test experience. "Tell me about a time you found an error in a payroll run after approval — what did you do?" tells you far more than "what would you do if…".

Probe until you have the specifics: the situation, what they personally did as opposed to what the team did, and how it ended.

Score against a defined scale, during the interview

Write the scale before you start, with a description of what a weak, adequate and strong answer contains for each criterion. Score as you go, not afterwards — memory reorganises itself around the overall impression within minutes of the candidate leaving.

A simple 1–5 with written anchors is enough. The point is not precision. The point is that two interviewers scoring the same answer differently have to explain why, and that conversation is where bias gets caught.

Use more than one interviewer, and score separately first

Panels work only if members score independently before discussing. Otherwise the most senior or most confident voice sets the tone and the rest converge on it.

Score alone, then compare. Where scores diverge sharply, that is the most useful conversation of the whole process.

Test the work where you can

For most roles a short, realistic work sample predicts performance better than any interview question. Give an accounts candidate a reconciliation with an error in it. Give an HR candidate a grievance letter to draft a response to.

Keep it to something that can be done in under an hour, make it relevant, and never ask candidates to do real unpaid work.

Guard against the usual biases

  • Similarity bias: rating people higher because they resemble you in background or manner
  • Halo effect: one strong answer or an impressive employer colouring every other rating
  • Recency: the last candidate of the day being remembered most clearly

Structured scoring blunts all three, because each criterion is rated on its own evidence rather than on a general feeling.

Keep the record

Scores, notes and the reason for the decision, kept for every candidate. It is what allows you to explain a decision if it is ever questioned, and it is also how you learn: comparing interview scores against performance a year later is the only way to find out whether your process predicts anything at all.

What it costs and what it returns

Building the criteria, questions and scale for a role takes an afternoon. It is reusable for every future hire into that role, it cuts interview time because the conversation has a spine, and it substantially improves the chance that the person you appoint is the right one.