Even the best single method — a structured interview at .42 — leaves most of the variation in performance unexplained. The way to predict better is to combine methods that measure different things, so each adds information the other misses. That added prediction is called incremental validity.

The classic evidence. In Schmidt & Hunter (1998), pairing a cognitive-ability test with a second method produced the strongest composites: GMA plus a work-sample test and GMA plus a structured interview each reached a composite validity of about .63, and GMA plus an integrity test about .65. The exact magnitudes are pre-2022 and now too high (cognitive ability’s own validity was revised down to .31), but the principle holds and is reinforced by the newer work: Sackett et al. (2023) show that well-chosen composites still substantially out-predict any single method, and that — because cognitive ability no longer dominates — interviews, work samples, biodata, and knowledge tests can be combined to reach strong validity without leaning on a cognitive test at all.

Why combining works. A structured interview captures interpersonal and judgement signals; a work sample or knowledge test captures task skill; a cognitive test captures learning speed. They correlate only modestly with each other, so each adds something. Two methods that measure the same thing add little; two that measure different things add a lot.

The cost of the interview-only hire. Most small employers decide on a single unstructured interview. That is the weakest defensible basis for a hire: unstructured-interview validity is about .19 (Sackett et al., 2022), it is the most bias-prone method, and it gives you no second, independent read on the candidate. Upgrading to a structured interview roughly doubles its validity; adding one more job-relevant method (a work sample for experienced hires, a knowledge test, or a structured reference verification) is the single biggest improvement most SMBs can make.

A sensible SMB stack. Structured interview + one job-relevant test (work sample or knowledge), with references used only to verify and red-flag at the end, as set out in the reference-checks note. This is cheap, high-validity, and — because every component is job-related and consistently applied — more defensible than a gut-feel conversation.

For an Ontario SMB, combining methods is also a risk-management move: a documented, multi-method, job-related process is both more accurate and easier to defend if a hiring decision is ever challenged under the rules in the Compliance cluster. The downside you are insuring against is quantified in the numbers cluster.

Source: Schmidt & Hunter, "The validity and utility of selection methods," Psychological Bulletin, 1998 ·

Last reviewed .

Confidence: Industry consensus