Candidate Assessment Software: Testing Capability, Not CVs

The numbers behind this

What is candidate assessment software, and where does it sit?

After screening and before the final decision. Screening answers whether somebody meets the stated requirements, and it works on inbound applications for one live role. Assessment answers how capably the shortlist can do the job, using exercises, tests and scored interviews. The two get sold together and are frequently the same product, but they fail in opposite directions, so keep them separate in your head. Over-screen and you lose capable people on a keyword. Over-assess and you lose them to your own process. The boundary matters between these pages too: filtering CVs and applications belongs to resume screening software, while everything here concerns measuring capability once a person is already worth your time. If you are still cutting four hundred applicants down to twelve, that is screening, and this page will not help much.

Which assessment methods are worth using, and for which roles?

The ones closest to the actual job. A short work sample built from a real task somebody would face in their first month tells you more than an hour of conversation about hypotheticals. Structured interviews come close behind and cost far less to run, provided every candidate gets the same questions and each answer is scored against a defined scale. Cognitive and situational judgement tests earn their place at volume, where applicants outnumber anyone's capacity to interview and consistency is the entire point. Personality questionnaires are the most oversold category in hiring. They can support a development conversation after somebody joins, and they are a weak basis for rejecting anybody. Match the instrument to the stakes, because a high-volume support role and a head of engineering warrant completely different treatment. A shared bank of structured interview questions does more good than most tools.

How do you build a work sample that respects a candidate's time?

Cap it, pay for it when it runs long, and make it a slice of the job rather than a project. Ninety minutes is a fair ceiling for most roles, and anything past half a day should either be paid or replaced with a live exercise. Take a genuine task from the last month, strip out the context that would take a new joiner a week to acquire, and give every candidate the identical brief. Say up front how long it should take, what you are assessing, and when they will hear back. Then score it against those criteria instead of reacting to it. Three things ruin an otherwise good exercise: unpaid work that looks suspiciously like a deliverable, a brief so vague that people spend their time guessing at intent, and a panel that marks on presentation. Give it to an existing employee first and watch how they get on.

What makes a scorecard produce a decision instead of an opinion?

Anchors, independence, and a small number of criteria. A scorecard with ten dimensions on a one-to-five scale produces noise, because nobody can hold ten judgements apart in an hour. Choose four or five things that genuinely predict success in the role, write a short description of what a one, a three and a five look like on each, and have every interviewer submit their scores before the debrief opens. That last step matters more than the form design does. Once the most senior person in the room speaks first, everybody else quietly converges on their view, and the panel produces a single opinion wearing five hats. Require evidence beside each score, meaning what the candidate said or did rather than how they came across. Then debrief on the disagreements, since that is where the information lives. Scorecards are among the parts of an ATS worth testing properly.

What should you ask a vendor about validity and adverse impact?

Ask what the test predicts, which populations it was studied in, and what evidence exists that scores relate to performance in roles resembling yours. Validity is the plain question of whether a score means anything for the job at hand, and a test can be carefully built and still irrelevant to your role. Adverse impact describes a selection step passing candidates from one group at a materially lower rate than another, which is why pass rates are worth measuring stage by stage instead of only at offer. A vendor who cannot describe their evidence, or who answers with customer anecdotes instead, has told you something useful. Obligations around testing, accommodations, data handling and record-keeping vary by jurisdiction and change often. None of this is legal advice, so have counsel review any assessment you make a condition of progressing.

When does an assessment predict performance, and when is it just a filter?

It predicts when the thing being measured is genuinely required by the job and the format lets a capable person demonstrate it. It filters when it measures something correlated with background rather than ability: familiarity with one specific tool the role would train anybody on, speed against an arbitrary timer, or fluency in an idiom of English the work never requires. The distinction is testable rather than philosophical. Look at whether high scorers went on to perform once hired, and whether people who fail are failing on the skill or on the format. Watch who drops out as well as who fails, since an exercise that quietly loses carers and people working a notice period is filtering on availability. One useful discipline: ask what decision changes based on this score. If the honest answer is nothing, delete the stage.

How much assessment is too much?

When the process costs more, in candidate time and withdrawals, than the risk it removes. Every extra stage loses people, and those with options leave first, so an over-engineered process systematically retains the candidates who have the fewest alternatives. Set a budget in hours per candidate before designing anything: one to two hours in total for a volume role, perhaps four for a senior individual contributor, more only for executive hires where a mistake is expensive to unwind. Then fit the assessment inside that budget rather than bolting on a round every time somebody feels nervous. Watch the calendar as well as the count, because two stages nine days apart make a longer process than four stages run inside a week. Compressing the calendar is usually the cheaper fix, and interview scheduling is where most of that time hides.

Assessment methods: what each measures, what it costs the candidate, and where it fits