A pre-employment assessment is any structured test used to gather evidence about a candidate before you hire them. That covers work samples, skills tests, cognitive ability tests, personality and behavioural questionnaires, structured interviews and reference checks. They are not interchangeable. Each answers a different question, and the most common mistake is running one because it is available rather than because you have a question it answers.
Start from the question. If you cannot say what a result would change, do not run the test.
The five questions, and what answers each
Can they do the work? A work sample or skills test. The strongest evidence available, because it is the job in miniature rather than a proxy for it.
Can they learn work they have not done? A cognitive ability test. This is what you want when you are hiring for potential, into a role the person has not held, or into something that will change underneath them.
How will they work with other people? A behavioural or personality assessment, and a structured interview. Weaker predictors on their own, useful when tied to specific behaviours the role genuinely demands.
How do they think through a problem? A structured interview with real follow-ups, or a case exercise.
Have they actually done what they said? A reference check, done properly, with specific questions rather than confirmation of dates.
Most roles need two of these. Very few need four, and stacking assessments has a real cost in candidate drop-off.
What actually predicts performance
The research on selection methods is unusually mature, and the broad ordering has held up across decades of meta-analysis.
Work samples and structured interviews are consistently at or near the top. Cognitive ability tests are strong, and are among the better predictors for roles involving complex or novel work. Personality measures are weaker on their own, with conscientiousness the most consistently useful factor. Unstructured interviews are near the bottom, which is awkward given they remain the most-used method in the world. Years of experience and education level are weak predictors of performance once someone clears a basic threshold.
Two practical conclusions follow.
First, combining methods beats any single method, because they are measuring different things. A work sample plus a structured interview is a strong, cheap combination for most roles.
Second, the method you already run is probably your weakest one. Almost every employer interviews. Comparatively few structure it. Fixing the interview costs nothing and improves your prediction more than adding a new test.
Matching the assessment to the role
High-volume, entry-level. A short cognitive or situational test early, to sort a large field consistently. The alternative at volume is screening on proxies like education, which predicts worse and excludes more.
Skilled and technical. A work sample. Nothing else comes close, and a short, well-defined task beats a long take-home that only candidates with free evenings can complete.
Professional and specialist. A structured interview plus a targeted skills test on the one or two capabilities that genuinely differentiate.
Management. A structured behavioural interview against defined leadership behaviours, plus references that ask specific questions about how they actually managed people.
Executive. Depth over breadth. A rigorous structured process, real referencing, and assessment against the specific mandate rather than a generic leadership profile.
What assessment costs
Per-candidate pricing is normal, which means the real cost is the per-candidate price multiplied by how many candidates you put through, and the second number is the one people forget.
HRM8's Assessment Hub is priced per candidate with no subscription: a single library skill from US$25, a cognitive aptitude test at US$30, a behavioural assessment at US$45, a leadership assessment at US$60, an AI interview at US$25 or US$145 for a job pack of ten, and a reference check at US$75, with a job pack of three at US$165. Curated role assessments run from US$35 for entry level to US$130 for executive roles.
Credits are US$5 each with volume discounts from ten to thirty percent, and they do not expire, which matters more than it sounds for employers whose hiring is lumpy rather than steady. Full detail is on the Assessment Hub page.
The economics are straightforward. If assessment costs you a few hundred dollars across a shortlist and prevents one wrong hire a year, it has paid for itself several times over. The cost that should worry you is not the fee, it is the drop-off from asking candidates to complete too much.
Where assessment goes wrong
Testing for its own sake. If you have not decided what a given score changes, you have added a stage and gathered no decision. Agree the rule first: what result progresses, what result is a flag to explore, what result stops the process.
Too much, too early. A ninety-minute assessment before a human has spoken to the candidate is a good way to lose exactly the people who have options. Put light, short assessment early and depth late.
Using a test outside what it measures. A personality questionnaire is not an integrity test. A cognitive test is not a culture assessment. Tests used outside their validated purpose produce confident numbers that mean nothing, and they are indefensible if challenged.
Treating the score as the decision. It is one input. A candidate who scores well on everything and cannot explain their own work is telling you something the score did not.
No accessibility adjustments. Timed tests disadvantage candidates with dyslexia, ADHD and various disabilities in ways unrelated to job performance. Offer adjustments, say so up front, and make asking for one straightforward and private.
The legal position has moved, and it is still moving
This is now a live compliance area rather than a theoretical one, and the direction is consistent: if an automated system contributes to an employment decision, you are expected to be able to explain it and to have checked it for bias.
The EU AI Act classifies AI systems used in recruitment and selection as high risk, bringing obligations around risk management, data governance, transparency and human oversight. New York City's Local Law 144 requires an annual independent bias audit of automated employment decision tools, publication of the results, and advance notice to candidates. Colorado has passed broad AI legislation covering consequential decisions including employment. In Australia, automated decision-making sits under Privacy Act obligations. UK regulators have issued guidance on AI in recruitment.
The practical implications for any employer, regardless of jurisdiction:
HRM8's position on this is explicit. The Assessment Hub is AI-scored and never auto-rejects a candidate. Any disqualifying result routes to the employer to sign off, and the employer owns the hiring decision. The Hub is designed against the major AI governance frameworks. What it does not do is make the decision for you, and that is deliberate.
- Know what your tools do. If a vendor cannot explain how a score is produced, you cannot explain it either, and that is now your problem rather than theirs.
- Keep a human on every consequential decision. Narrowing and ordering by system is defensible. Automatic rejection is where the exposure concentrates.
- Tell candidates. Disclosure is becoming standard and is required in several jurisdictions.
- Keep records. What was assessed, what the result was, what decision followed and who made it.
A sensible default
For most roles, this is enough:
1. Screen on written criteria. 2. A short skills test or work sample, matched to the actual job. 3. A structured interview against a scorecard. 4. A reference check with specific questions, for the finalist.
That is defensible, it predicts well, it is affordable, and it does not ask candidates for hours of unpaid work. Add a cognitive test where you are genuinely hiring for potential, and a behavioural assessment where specific interpersonal demands are central rather than incidental.
Common questions
What is a pre-employment assessment?
Any structured test used to gather evidence about a candidate before hiring: work samples, skills tests, cognitive ability tests, behavioural questionnaires, structured interviews and reference checks.
Which assessment best predicts job performance?
Work samples and structured interviews are consistently the strongest, with cognitive ability tests close behind for complex or novel work. Combining two methods beats any one.
Are pre-employment tests legal?
Yes, when they are job-related, applied consistently, and accessible. The compliance burden rises sharply when the tool is automated and contributes to a decision. The EU AI Act treats recruitment AI as high risk, and New York City requires an annual bias audit and candidate notice for automated employment decision tools.
How much does candidate assessment cost?
Typically per candidate. HRM8's range runs from US$25 for a single skill or an AI interview to US$130 for an executive role assessment, with credits that do not expire.
Can an assessment reject a candidate automatically?
It should not. HRM8's Assessment Hub never auto-rejects. Results route to the employer, who owns the decision. Automated rejection is where the legal exposure concentrates.
Start from the question you cannot otherwise answer, pick the method that answers it, keep the battery short and early stages light, agree what each result changes before you run it, and keep a human on every consequential decision.