A skills taxonomy gives an organisation a more structured way to describe the skills its workforce needs. Adding proficiency levels makes that picture more useful. It is one thing to know that someone has a skill. It is another to know how well they can exercise it.

The difficulty is that a proficiency rating can look more certain than the evidence behind it.

Two employees might both be recorded as having “advanced” complaint-handling skills. For one, that rating may come from a self-assessment or completed course. For another, it may reflect a manager’s judgement after repeatedly observing them handle complex guest situations, exercise discretion and resolve problems without unnecessary escalation.

The label is the same. The basis for the claim is not.

That does not make taxonomies or proficiency models less useful. It does mean that the organisation needs to understand what a proficiency rating actually represents. A level assigned to a person is not simply a description. It is a claim about what that person can do.

Once skills data starts informing decisions about development, deployment, promotion or workforce readiness, the important question is no longer only what the rating says. It is how much evidence sits behind it and what that evidence is strong enough to support.

The evidence required depends on what the organisation wants to do with the claim

Not every skill claim needs the same level of evidence.

If the purpose is exploratory, relatively light evidence may be enough. A self-declared skill, an inferred skill or a manager’s view can still be useful for recommending learning, helping someone explore possible career paths or surfacing talent that may warrant a closer look. In those situations, the data is a signal. It does not need to carry the same evidentiary weight as a formal judgement about readiness.

The burden changes when the consequences of the decision increase.

Using a proficiency claim to decide whether someone should be promoted, deployed into a critical role, included in succession planning or treated as ready for independent work creates a different standard. The organisation is no longer using the data to explore a possibility. It is relying on the claim to make a decision that affects the person, the work or both.

This is consistent with established assessment practice. Validity is not simply a property of a test or rating in isolation. It also concerns whether the interpretation and use made from the evidence are justified. South African assessment practice similarly places emphasis on evidence being sufficient for the judgement being made.

A more useful question is what decision is this evidence strong enough to support?

That principle avoids an unnecessary choice between informal skills data and heavy assessment. Evidence can remain proportionate. But before deciding what evidence is enough, the organisation first needs to be clear about what a proficiency level is actually claiming.

Proficiency should describe how a skill is exercised, not merely attach a number to it

A proficiency model is only useful if its levels mean something in practice.

Labels such as beginner, intermediate, advanced or expert may create the appearance of precision, but they are difficult to use consistently unless the organisation can explain what changes from one level to the next. A stronger model describes differences in how a skill is exercised.

That may include dimensions such as complexity, autonomy, judgement, scope and responsibility. SFIA, for example, distinguishes levels partly through factors such as autonomy, influence and complexity rather than relying only on broad labels.

For example, consider a hotel supervisor responsible for handling guest complaints. A lower level of complaint-handling proficiency might mean resolving routine issues using established procedures and seeking guidance when the situation falls outside them. A higher level might mean independently handling more complex or emotionally charged complaints, exercising discretion within delegated authority, adapting when the standard response does not fit, and recognising when escalation is necessary.

That gives the proficiency level something concrete to describe.

It also exposes an important distinction. An organisation can decide that a particular role requires complaint handling at that higher level. That is a role-design decision. It is a statement about the work.

Saying that a particular supervisor actually operates at that level is different. That is a claim about the person, and it requires evidence.

Proficiency should therefore define the level of performance being claimed before the organisation decides whether the available evidence is strong enough to support it. It also establishes the limit of proficiency itself: even well-defined proficiency across individual skills does not automatically amount to broader capability.

Proficiency in individual skills is still not the same as capability

Proficiency describes how a particular skill is exercised. Capability asks a broader question: can the person bring the relevant skills, knowledge, behavioural repertoires and judgement together effectively across the situations the work requires?

Real work rarely arrives as a set of isolated skills. A hotel supervisor may be highly proficient in communication, problem solving and complaint handling when each is assessed separately. That does not yet establish that they can combine them effectively when a guest is angry, the usual remedy is unavailable and the situation requires both discretion and sound escalation judgement.

The evidentiary burden therefore changes as the claim becomes broader.

One successful demonstration shows that the person performed effectively once. Repeated success under similar conditions gives stronger evidence of consistency. But consistency is still not the same as range. A broader capability claim becomes more credible when effective performance can also be demonstrated across relevant variation in complexity, uncertainty or context.

This is where reliability and adaptability acquire an evidentiary meaning. Reliable performance means more than a single good result. Adaptive performance means more than repeating the same response successfully. It requires evidence that the person can adjust appropriately when the situation changes without losing effectiveness.

Assessment literature on sufficiency and generalisability points in the same direction: broader claims are better supported when evidence samples more than one task, occasion or context rather than relying on an isolated performance.

That does not mean every capability judgement requires an extensive assessment process. It means the breadth of the evidence should be proportionate to the breadth of the claim.

Once capability is understood in those terms, another question follows naturally: can one source of evidence reasonably show all of it?

Different evidence sources answer different questions

No single evidence source is likely to show every dimension of a broad capability claim equally well.

A knowledge assessment may show whether someone understands the relevant principles. A scenario or simulation may reveal how they reason through a difficult situation. A work sample can show applied skill. Manager observation may reveal patterns over time. Operational data can show what is happening in the work.

None is inherently better than the others in every context. They answer different questions, and each has limits.

This is where triangulation becomes useful. Sometimes several sources corroborate the same claim. At other times, they are complementary because each reveals a different part of the capability being judged. The value lies less in accumulating more evidence for its own sake than in combining evidence that exposes different strengths and blind spots.

Consider the hotel supervisor again. A simulation might show strong judgement in a complex service-recovery scenario. Repeated manager observations may suggest that the supervisor handles difficult guests well in practice. Yet operational data may still show slow resolution times.

That disagreement should not automatically be treated as evidence that one source is wrong. Different sources may be revealing different parts of what is happening.

Programmatic and multisource assessment approaches are useful here because they treat complex judgements as something built from multiple pieces of evidence rather than reduced to a single score.

When evidence sources diverge, the task is not simply to choose the one we trust most. It is to understand what each source is actually telling us. Operational performance, in particular, requires further interpretation.

Performance is powerful evidence, but it still needs interpretation

Workplace performance is often among the strongest evidence available because it shows what actually happened in the conditions of real work. But that strength can also make it easy to overinterpret.

Performance evidence tells us what someone did under particular conditions. Capability evidence supports a broader inference about what that person can reliably do across the relevant range of conditions. They are related, but they are not the same.

A hotel supervisor may consistently show sound judgement and strong service-recovery capability, yet still have poor complaint-resolution times because compensation approvals are tightly restricted, the property-management system is slow, or other teams delay the final resolution. In that case, weak operational performance does not necessarily indicate weak capability.

The reverse can also be true. Strong resolution metrics may reflect highly scripted processes, generous automatic remedies or close managerial support. The outcome may be good, but the evidence tells us less about how independently the supervisor can diagnose a difficult situation, exercise judgement or adapt when the standard process no longer fits.

This is consistent with performance-improvement thinking more broadly. Frameworks such as COM-B distinguish capability from the opportunities and conditions that influence whether behaviour can occur, while systems-oriented approaches such as Rummler-Brache place individual performance within a wider work environment.

Operational data remains important, often indispensable. It simply needs to be interpreted in context.

If performance is treated as capability without considering the conditions that produced it, organisations risk diagnosing the person when the constraint sits elsewhere, or attributing capability to the person when the system is doing much of the work.

A useful skills model should preserve how much confidence its claims deserve

If skills data is going to support consequential workforce decisions, the organisation needs some visibility into the basis of the claims it is relying on.

That does not mean building a formal assessment regime around every skill in the taxonomy. In many cases, lighter evidence will remain entirely appropriate. But a self-declared skill, a manager-assigned proficiency level and a capability demonstrated repeatedly across varied situations should not appear as though they carry the same evidentiary weight.

A more useful skills model should preserve at least some of those differences.

That might include information about the source of the claim, when it was last evidenced, how it was validated, or the degree of confidence attached to it. SFIA assessment guidance and South African competence-assessment practice provide useful precedents for linking claims to supporting evidence, even if they do not prescribe this exact organisational model.

We can think of this as an evidence-aware skills model, or more broadly as an evidence architecture around the skills model. These are useful working terms rather than established industry frameworks.

The aim is not certainty. It is to make the basis of the claim, and any uncertainty around it, more visible.

That brings the article back to the practical test established earlier: what decisions is this evidence good enough to support?

A useful skills model should help the organisation answer that question rather than hiding materially different evidence behind the same proficiency label.

The value of a skills model depends partly on knowing where its evidence stops

A taxonomy gives the organisation a shared way to describe skills. A proficiency model makes those descriptions more specific by defining the level at which a skill is expected to be exercised. Evidence determines how confidently that level can be assigned to a person and what decisions it can reasonably support.

A better way to think about skills data is as a set of claims with different levels of evidentiary support, rather than a set of uniform workforce facts.

Not every claim needs extensive validation. For exploratory or developmental purposes, lighter evidence may be entirely sufficient. But as the consequences of the decision increase, the evidence should become stronger and more representative of the capability being claimed.

The value of the model therefore depends partly on knowing where its evidence stops. The stronger the workforce claim or the more consequential the decision, the stronger the evidence needed to support it.

References

Briesch, A.M., Swaminathan, H., Welsh, M. & Chafouleas, S.M. (2014). Generalizability theory: A practical guide to study design, implementation, and interpretation. Journal of School Psychology, 52(1), 13–35.

Fraser, S.W. & Greenhalgh, T. (2001). Coping with complexity: educating for capability. BMJ, 323, 799–803.

Kane, M.T. (2013). Validating the interpretations and uses of test scores. Journal of Educational Measurement, 50(1), 1–73.

Michie, S., van Stralen, M.M. & West, R. (2011). The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implementation Science, 6, 42.

Rummler, G.A. & Brache, A.P. (2012). Improving Performance: How to Manage the White Space on the Organization Chart. 3rd ed. San Francisco: Jossey-Bass.

South African Qualifications Authority (SAQA). (2017). National Policy and Criteria for Designing and Implementing Assessment for NQF Qualifications and Part-Qualifications and Professional Designations in South Africa. Pretoria: SAQA.

SFIA Foundation. (n.d.). Assessor’s guidelines. Accessed 6 November 2026.

SFIA Foundation. (n.d.). How SFIA works: Levels of responsibility and skills. Accessed 6 November 2026.

van der Vleuten, C.P.M., Schuwirth, L.W.T., Driessen, E.W., Dijkstra, J., Tigelaar, D., Baartman, L.K.J. & van Tartwijk, J. (2012). A model for programmatic assessment fit for purpose. Medical Teacher, 34(3), 205–214.