The science
What actually predicts whether someone can do the job
Predictive validity, across two major meta-analyses shown together. The longer the bar, the better that signal predicts real job performance.
Work sample testsSackett 2022 — Schmidt & Hunter 1998.33–.54
Structured interviewsSackett 2022 — Schmidt & Hunter 1998.42–.51
Years of experienceSchmidt & Hunter 1998.18
Years of educationSchmidt & Hunter 1998.10
Scale 0 to 0.6
The 2022 revision (Sackett et al.) lowered most estimates across the board, including credential-based methods — but the gap between demonstrated work and claimed credentials holds under both analyses.
Simulations decide who gets interviewed; structured interviews decide who gets hired. Luminid replaces resume screening, not the interview.
Predictive validity coefficients of selection methods. Schmidt & Hunter 1998, and the Sackett et al. 2022 revision.| Selection method | Predictive validity coefficient |
|---|
| Work sample tests | 0.33–0.54 |
| Structured interviews | 0.42–0.51 |
| Years of experience | 0.18 |
| Years of education | 0.10 |