Skills-first hiring is not a keyword exercise: how to collect evidence that deserves trust

Removing a degree requirement is easy. Replacing it with a better decision is not.

That gap explains why skills-first hiring can become little more than new wording in a job advert. The organisation removes a credential filter, adds a list of twenty skills and then interviews candidates exactly as before. The World Economic Forums Future of Jobs Report 2025 found that skills gaps remain a major barrier for employers and that skills assessments are expected to play a substantial role in hiring. OECD work also highlights both the opportunity and the implementation difficulty of skills-first approaches. The promise is real, but only if skills become evidence rather than keywords.

Define the outcomes of the role

Begin with what a successful person must produce, decide or improve. Ask managers and experienced workers for recent examples. What did good performance look like? What made a task difficult? Which errors mattered? Which capabilities were learned after joining?

Convert the answers into a small number of observable requirements. Stakeholder management might become identifying competing needs, explaining trade-offs and securing a workable commitment. Data literacy might mean checking data quality, selecting an appropriate analysis and communicating uncertainty. The behavioural description gives assessment designers something they can actually observe.

Separate entry requirements from future potential. Employers often ask candidates to arrive with everything the current team has accumulated over years. That narrows the pool and turns experience into a hidden credential. Identify what must be demonstrated now, what can be supported through onboarding and what signals capacity to learn.

Remove proxies deliberately

A degree, previous title or years-of-experience rule may sometimes reflect a real requirement, particularly in regulated roles. Often it is simply an inherited shortcut. For each proxy, ask what capability it is meant to indicate, whether the connection is supported and whether a direct evidence source is practical.

Do not remove the visible requirement while preserving it elsewhere. A recruiter who still favours familiar universities, a hiring manager who expects an identical title and an applicant system that ranks career continuity can recreate the same exclusion. Audit the complete funnel and the data used by any ranking technology.

Be candid about trade-offs. Direct assessment takes candidate time and organisational effort. The answer is not to test everyone extensively. Use a light initial screen grounded in essential requirements, then gather richer evidence only from a smaller group.

Choose evidence that resembles the work

A good work sample contains the judgement of the role without requiring free labour or insider knowledge. Give a customer-success candidate a short, fictional account with competing priorities and ask for a response plan. Ask a data candidate to identify flaws in a small dataset and explain what they would do next. Let a manager review a scenario and conduct a structured conversation.

Keep the exercise bounded. Candidates should know the expected time, available tools and scoring criteria. Avoid assignments that could be used commercially. Offer an alternative where the format creates an access barrier. Pilot the task with people unfamiliar with your internal language; otherwise you may be measuring knowledge of your organisation rather than capability.

Not every skill needs a simulation. Structured interviews, portfolios, verified credentials, prior work examples and references can provide complementary evidence. Use several relevant signals when the consequence warrants it, but do not average incompatible measures into an unexplained score.

Make scoring structured and visible

Write a rubric describing observable differences between levels. Avoid labels such as excellent communication without examples. For the customer scenario, criteria might cover identification of needs, prioritisation, handling uncertainty and clarity of next steps. Weight criteria according to the role rather than what is easiest to score.

Give assessors representative examples and calibrate them before live hiring. Ask them to score independently, compare reasoning and resolve ambiguous criteria. Repeat a small calibration during longer campaigns to detect drift. Record the evidence and rationale beside the decision.

Structure should extend to interviews. Ask candidates the same core questions, use defined follow-ups and score before open discussion. This does not make the conversation robotic. It protects the evidence from being replaced by rapport, similarity or memory.

Respect the candidates opportunity to demonstrate

Skills-first hiring should widen opportunity, not exchange a credential barrier for an assessment obstacle course. Publish the stages, likely time and purpose. Do not ask candidates to complete multiple long tests before a human has reviewed basic eligibility. Provide a practice item and explain permitted assistance.

Build accessibility in from the beginning. Timers, custom interactions, video requirements and remote monitoring can introduce barriers unrelated to the skill. Make adjustment requests straightforward and confidential. If a capability can be demonstrated through more than one equivalent route, offer the choice.

Feedback is valuable, especially for later-stage candidates, but it must be accurate and safe. A generic automated paragraph can do more harm than a concise explanation of criteria and outcome. If AI drafts feedback, keep human approval and source evidence visible.

Use credentials as evidence, not destiny

Microcredentials and certificates vary widely. Ask what was taught, what was actually assessed, under what conditions, by whom and when. A verified practical credential may be strong evidence for one capability and irrelevant to another. A degree may provide useful broad evidence without proving current performance in a specific tool.

Let candidates combine evidence. Someone might present a certificate, portfolio and brief work sample. Preserve provenance so reviewers know which claim came from which source. Avoid turning diverse evidence into a single opaque skills score. A profile that shows strength, uncertainty and missing evidence supports a better conversation.

Connect hiring evidence to later performance

A selection process should learn from what happens after hiring. With appropriate privacy controls, compare assessment evidence with onboarding progress, role performance and retention. Do the criteria predict the outcomes they were intended to support? Are some items adding burden without useful information? Do certain groups encounter barriers that do not appear in final scores?

Do not use performance ratings uncritically as ground truth; they can contain their own bias and inconsistency. Combine manager evidence with work outcomes, structured observations and employee experience. Review the assessment when the role changes. A task based on last years workflow can quickly become a test of obsolete practice.

A realistic first pilot

1. Select one recurring role with enough hiring volume to learn.
2. Identify three to five essential entry capabilities from real work.
3. Remove one unjustified proxy from the job and screening process.
4. Create a short work-relevant exercise and a structured interview.
5. Build a rubric, calibrate assessors and test accessibility.
6. Explain the process and time commitment to candidates.
7. Monitor completion, subgroup outcomes, overrides and candidate feedback.
8. Compare evidence with later performance and revise the design.

Small employers can do this without a large transformation programme. A shared rubric, a thirty-minute work sample and a disciplined review can be more skills-first than an expensive platform attached to a vague competency model.

Write the candidate explanation before launch

Try explaining the selection process to a capable applicant who lacks the traditional credential. What will they be asked to demonstrate, why were those capabilities chosen, how will evidence be scored and who will make the decision? If the explanation collapses into our algorithm identifies potential, the process is not yet evidence-led.

A clear explanation improves design. It forces the team to remove criteria that have no defensible purpose and makes appeals easier to handle. It also lets candidates prepare for the nature of the exercise without giving away secure content. Publish enough detail to create a fair opportunity: expected time, format, tools, adjustment route and the broad criteria.

The standard is better evidence

Skills-first hiring is not achieved when the word degree disappears. It is achieved when a wider range of people has a fair opportunity to demonstrate relevant capability and the employer can explain why the evidence supports the decision.

Keep the role definition small and current, use proportionate work evidence, structure human judgement and learn from outcomes. Credentials can remain one signal, but they no longer have to carry meaning they were never designed to prove.