Proof of fit: what to collect before the first interview

• Collect four things before the first interview: a work sample tied to the job's main task, an answer to one question only a practitioner can answer, the numbers describing what the candidate owned, and one reference point you can check. Gathered at application, those four let you rank a hundred people on what they have shown you. They also give the interview somewhere to start other than a walk through the CV.

Why the CV cannot carry this

• A CV is a claim, written by the candidate, formatted for a machine that may read it badly. Harvard's hidden workers research counted 27 million people in the United States who want work and are screened out before anyone reads about them, and found that roughly 99 percent of the largest companies put applications through automated screening first. Ashby measured 291 applications per hire in the first quarter of 2026.

• At that volume the CV gets ten seconds. Ten seconds against a claim tells you almost nothing. Ten seconds against a work sample link, a scope number and a written answer tells you whether to spend the next five minutes.

• The argument for ranking on shown work rather than described work is set out in [hiring on proof of fit](/spaces/hiring-on-proof-of-fit).

The checklist

• Fifteen minutes to an hour of candidate time in total. Everything on that list can be captured in structured fields on the application form, which also means it survives a bad CV parse.

Build the work sample from a real problem

• Take something the team actually handled in the last quarter, strip the confidential detail, and cut it until it fits an hour. A support role gets three real tickets and a request for the replies. A finance role gets a messy sheet and one reconciliation question. A designer gets a screen that tests badly and a brief to fix it.

• Two rules keep this honest. Say the time limit in the invitation and mean it, because unpaid hours fall hardest on candidates with jobs and caring responsibilities. Accept prior work in place of the task, since a senior candidate with a portfolio has already produced the evidence you are asking for.

Write one question only a practitioner can answer

• The generic version, describe a challenge you overcame, can be answered by anyone with a browser. The useful version is narrow enough that it needs the actual job in it.

• For a payroll specialist: describe the last time a pay run was wrong after cut-off and what you did in the following 48 hours. For a field engineer: what do you check first when a site reports intermittent failures that clear before you arrive.

• A question that narrow is hard to answer from a browser, and the answers give the structured interview somewhere concrete to start.

What the research supports

• The reanalysis of personnel selection validity by Sackett and colleagues, summarised by the Society for Industrial and Organizational Psychology, revised the ranking most hiring teams learned. Structured interviews come out highest at .42, followed by job knowledge tests at .40, empirically keyed biodata at .38, work samples at .33 and cognitive ability tests at .31.

• The structured interview earns the top spot and it is expensive, so it belongs to the shortlist. The pre-interview evidence exists to decide who gets one. Job knowledge and a task sample are cheap to review at volume and they are about the work itself, which is what makes them the right filter before the expensive conversation.

Ask everyone for the same thing

• One set of requirements, one rubric, written before the first application arrives. Score each item on a 0 to 2 scale and keep the scores on the record.

• This is what makes the process defensible when someone asks why a candidate did not progress, and it is also what makes your screen improvable. A rubric that produces the same score from two reviewers is one you can adjust once you find out where it was wrong.

• How the scored list becomes a shortlist is covered in [what is candidate shortlisting](/spaces/what-is-candidate-shortlisting), and the operational version, evidence to shortlist inside a day, is in [how to build a shortlist in hours](/spaces/build-shortlist-in-hours).

Use the evidence in the interview

• The most common waste in a first interview is asking someone to summarise what they already submitted. Read the work sample first, then open with what you found in it.

• Questions worth asking: why they made a specific choice in the sample, what they would change with another day, which part of it they have done before under real pressure. You end up with an hour that produces new information, and the candidate leaves having talked about work rather than about their CV.

Where the evidence goes in the pipeline

• Collect it at application for roles where the volume is high and the task is well defined. Collect it after a short screening call for senior or scarce roles, where asking for an hour before any human contact costs you candidates you cannot replace.

• Either way the evidence arrives before the first interview, attached to a score, on a list that keeps every applicant in order. The same list logic is described in [screening hundreds of applications without a keyword filter](/spaces/screen-applications-without-keyword-filter).

• Start with one role. Write the four items, put them on the form, and compare the resulting shortlist against the one your current screen would have produced.