Using the Most Impactful Evidence for Evidence-Based Selection

An inside account of overhauling a hiring process at a large manufacturing organization details how skills-based, evidence-based selection replaced reliance on resumes, pedigree, and personal connections. The problem started with a simple discovery: when every hiring leader and recruiter was asked what their hiring process actually was, no two answers matched. Without a standardized process, there was no way to create equity in who got hired.

The fix rested on a few core ideas:

• Job families: grouping roles by the actual competencies and skills they require, rather than treating every position as unique.
• Predictive, reliable tools: selection instruments chosen because they correlate with actual job performance, not because they've always been used.
• Compensatory scoring: evaluating candidates on a fuller picture of relevant factors rather than a single pass/fail gate or a resume line.
• Clean data: recruiting teams trained to keep information accurate, since flawed inputs produce flawed hiring decisions.
• Ongoing outcome review: checking results for adverse impact and retention after the process launches, since a system that feels right on paper can still fail in practice.

Competencies get split into two categories: occupational competencies, which reflect the ability to perform a specific task and can be judged through experience or credentials, and foundational competencies, the underlying skills needed to succeed at work generally, often measured through assessments.

Research on selection methods points to general mental ability and cognitive assessments as the strongest predictors of job performance, followed by structured interview questions, with personality and career-interest measures showing promise as well. A free tool built on labor-market data offers a starting point for identifying which competencies matter for a given job function, though internal validation with actual top performers matters just as much as any external framework, since stakeholders will always insist their roles and their people are the exception.

Several real examples come up along the way: an assessment used as a gatekeeping test that got scrapped once data showed it didn't predict floor performance, a coding assessment for engineers that was simplified from a difficult test to a lighter quiz without sacrificing retention gains, and a tiered assessment system built separately for hourly, individual-contributor, and executive-level hiring.