Skills-Based Hiring Is Easy to Announce and Hard to Measure

Removing formal education requirements from job postings has become a widely adopted policy, supported by a reasonable argument: for a large share of roles, a degree functions as a proxy for capability that is both loose and correlated with background rather than ability, and dropping it enlarges the candidate pool at no cost to quality. The policy change is trivial to implement — an edit to a template — and the resulting increase in applications is immediate and easy to report. The difficulty appears one step later. A credential requirement, whatever its flaws, did a specific job in the process: it reduced a large applicant pool cheaply and gave a screener a defensible reason to advance one candidate over another. Removing it without replacing that function leaves the screening step with more volume and less structure, and screening under those conditions falls back on whatever informal signals are available — familiar employer names, referral source, how a CV reads. Those signals are frequently more biased than the credential that was removed, and unlike the credential they are not written down anywhere that can be audited. Building the replacement is real work. It means defining, for each role, what a person must actually be able to do; designing an assessment that measures it in a reasonable amount of time without becoming an unpaid work sample; validating that the assessment predicts on-the-job performance rather than test-taking; and training interviewers to use structured criteria consistently. Organisations that have done this report better outcomes, and they also report that it took years and dedicated resources — which is why announcements outnumber implementations by a wide margin. The honest test of whether a skills-based policy is real is not the wording of the posting but the composition of recent hires and the retention of those hires against people from conventional paths. If neither number moved, the requirement did not disappear; it moved from a stated criterion into an unstated one, where it is harder to challenge and impossible to measure. A filter you cannot see is not an improvement on a filter you can.