Stockertim Dictionary · Human resources
Validation study
Also called validity study, test validation
In short
Research showing that scores from a selection tool are related to real job performance or another relevant work outcome, for the jobs and people it is used on.
- Advanced
- Process
- Checked
In detail
A validation study tests whether a selection method does what it claims. The US Uniform Guidelines on Employee Selection Procedures recognise three kinds of evidence: criterion-related validity, which shows scores predict job performance or other important work behaviours; content validity, which shows the test samples the actual content of the job; and construct validity, which shows the test measures a quality that the job requires.
Studies need care. The outcome measure must itself be sound, the sample must be large enough and similar to the people the tool will assess, and results should be checked by group, because a tool can predict well for one group and poorly for another. Evidence from another employer or another job family transfers only when the jobs are genuinely comparable, which the guidelines address in their rules on transportability.
For AI tools, buyers ask vendors for their validation reports, read the population and jobs studied, and run a local study or pilot where possible. A tool without validation evidence is difficult to defend if its outcomes are challenged as discriminatory, in the United States and elsewhere.
Worked example
A made-up contact centre, Clearline, compares assessment scores with six-month quality ratings for 200 hires.
- Hires in the top score band: 50
- Of those, rated good or better: 40 / 50 = 0.8
- Hires in the bottom band rated good or better: 20 / 50 = 0.4
Top-band hires succeed twice as often, evidence of criterion-related validity that is then checked by group.
In practice
HR and procurement teams request technical manuals with validation evidence, check that the jobs and populations studied match their own, and record their validity review in the AI use case register.
Watch out
- Validity for one job or population does not transfer automatically to another.
- A predictive tool can still have adverse impact; validity and fairness are tested separately.
Learn it properly
Official sources
- 29 CFR 1607.5: General standards for validity studies · Legal Information Institute (US Code of Federal Regulations)
- 29 CFR 1607.14: Technical standards for validity studies · Legal Information Institute (US Code of Federal Regulations)
Facts last checked on 3 November 2026. Rules and figures can change: check the official source before you rely on them.
Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.