What candidate assessment actually is, and how to choose a method

Candidate assessment is the practice of measuring what someone can do before you hire them, rather than inferring it from what they say they have done. It covers any structured method that produces evidence you can compare across people: a work sample, a structured interview, a psychometric test, a reference check.

It is not the same as screening, though the words get used interchangeably. Screening removes people — it answers "is this candidate worth an hour of my time?" Assessment ranks the ones who are left, and answers a harder question: "which of these people will do the job well?" A process can screen heavily and assess not at all, which is the most common failure mode in small teams.

The distinction matters because the two have opposite error costs. A screen that is slightly too harsh loses you a good candidate quietly. An assessment that is slightly wrong puts the wrong person in the job for a year.

Almost every assessment in use is one of four things, or a combination. They are usually presented as a menu of equivalent options. They are not equivalent — each measures something different, and each has a specific blind spot that the others cover.

Read the third column first. The reason most processes feel unsatisfying is that they run one method well and treat its blind spot as noise. A team that only runs interviews hires people who interview well. A team that only runs work samples hires people who can do the task and cannot explain it to a stakeholder six weeks later.

The useful question is not "which method is best" but "what does this role fail on?" Pick the method that measures the thing most likely to go wrong.

• Individual contributor, output you can look at — a work sample is the highest-signal hour you will spend. Engineering, design, writing, analysis. See assessing technical skill specifically for how to build one that is not a puzzle test.
• Roles where the output is a conversation — support, sales, account management, anything client-facing. A work sample measures the wrong half. Testing how someone explains their work is the harder and more useful screen.
• Management — no method is strong here, and the honest answer is that you are combining a structured interview with reference checks and accepting more uncertainty than you would like.
• High volume, low variance — hundreds of applicants for a role where the work is genuinely standardised. This is the one case where psychometric batteries earn their cost, and also the one carrying the most compliance weight.

Most roles want two methods, not four. A work sample and a structured interview covers the majority of hiring outside high-volume operations, because the interview covers the sample's blind spot directly: you ask them about the thing they just built.

Assessment is the third of five stages, and putting it anywhere else is what makes hiring processes feel long and produce weak decisions.

• Sourcing — getting candidates in front of you at all. Nothing here is assessment.
• Screening — a filter, not a ranking. CV, a short form, minimum requirements. Fast and slightly harsh.
• Assessment — the structured measurement. This is the stage that produces the evidence you will decide on.
• Decision — comparing scored candidates against the rubric you wrote before you saw anyone.
• Offer and close — no longer measurement, though candidates are still assessing you.

The common mistake is running assessment last, after several rounds of conversation, as a formality to confirm a decision already made. At that point it is not measurement — it is a rubber stamp with a cost attached, and everyone involved can feel that it is. Assessment belongs before the opinions form, not after.

A method without a rubric is not an assessment. It is a structured way of generating an impression, and impressions do not survive contact with a second interviewer, a week's gap, or a candidate who asks why they were rejected.

The rubric has to exist before you see the first candidate. Written afterwards, it will describe the person you already liked — which feels like rigour and is the opposite. Decide what good looks like for each thing you are measuring, decide what weak looks like, and write both down.

Score per competency rather than overall. A single number out of ten collapses the useful information: two candidates can score seven for opposite reasons, and the whole point of assessing was to know which is which. Strengths, gaps, and a confidence signal per answer will tell you more than any aggregate.

Assessment is a measurement taken under artificial conditions, in an hour, of a person who knows they are being measured. It is far better than the alternative and it is not the job.

• Whether they will still be motivated in month eight, when the work is maintenance rather than the interesting problem you screened on.
• How they behave when they disagree with someone more senior. No assessment format creates real stakes.
• Whether they will fit your team specifically, as opposed to a competent team in the abstract.
• Anything about growth. You are measuring a snapshot, and the people who compound fastest are frequently not the strongest on the day.

The practical response is to treat a strong assessment as a reason to hire, never as a guarantee, and to keep the probation period real. A process that produces certainty is not measuring anything difficult.

Nothing above needs a tool. Write one scoped task that mirrors the first week of the job. Write a rubric with three or four competencies and a description of weak and strong for each. Run the same task and the same rubric for every candidate. That is a complete assessment process and you can build it in an afternoon.

The reason teams skip it is not disagreement — it is that the afternoon has to happen again for the next role, and the fifth candidate arrives before the first rubric is finished. If you want the manual version in five steps, five checks you can run by hand is the short form. The time cost is the actual problem, and it is the only part of this a tool solves.