The Science Behind Evidence-Based Hiring
"Evidence-based" is a phrase that gets applied loosely across HR technology, worth actually tracing back to where the real evidence comes from, since the claim only means something if the underlying research holds up to scrutiny.
The foundational research
The most influential work in this area is a large-scale meta-analysis by Frank Schmidt and John Hunter, published in 1998, which pulled together decades of prior research on selection methods and calculated how strongly each one actually predicts job performance. This wasn't a single study or a vendor's internal data, it was a synthesis of a huge body of independent research, which is precisely why it's held up as a reference point for so long.
What the research actually found
The methods came out ranked by validity, a statistical measure of how strongly a method's results correlate with real, subsequent job performance. General mental ability tests and structured work-sample assessments ranked at the top. Unstructured interviews, the format still used by the overwhelming majority of employers, and years of job experience both ranked considerably lower. The gap between the best and worst methods wasn't small, some commonly used approaches were shown to add barely more predictive value than chance.
Independent confirmation, not just one source
This isn't a finding that rests on a single paper. The US Office of Personnel Management, the US federal government's own HR authority, independently confirms work-sample and simulation-based assessment as a high-validity selection method in its own published guidance, citing the same underlying research base. When a government personnel authority and an academic meta-analysis land on the same conclusion, independently, that's a genuinely strong signal the finding is robust, not a one-off result.
Why work samples specifically perform so well
The research offers a fairly intuitive explanation once you see it stated plainly: a work sample directly observes the behaviour you actually care about, how someone approaches a realistic task from the role, rather than inferring it indirectly from a conversation, a personality questionnaire, or a written claim. Indirect measures introduce more noise at every step removed from the real behaviour. A work sample removes several of those steps at once.
What this means practically
Any hiring method claiming to be "evidence-based" is worth holding to this actual standard: is it grounded in research showing it predicts real job performance, or does it just feel rigorous because it produces a score? A personality test produces a number too. Only some selection methods have genuine research behind the claim that the number means anything.
Sources: this piece references the meta-analysis by Frank L. Schmidt and John E. Hunter, "The Validity and Utility of Selection Methods in Personnel Psychology" (Psychological Bulletin, 1998), and selection-method guidance published by the US Office of Personnel Management.