Build an evidence-weighted capability portfolio

A transferable capability is not a skill label in a résumé. It is an action performed under conditions close enough to the target task, with a result that another person can inspect. A service supervisor may say she ‘analyzes data’ because she produces a weekly report. An adviser should distinguish compiling numbers, checking their source, explaining a change, and recommending a defensible response. These are different actions with different evidence requirements. Begin with a small portfolio of work samples rather than a long list of adjectives: a report, a decision record, a correction log, or a supervisor's observation. Obtain the person's permission and remove confidential or customer data before a sample enters a learning file. A sample description is not automatically proof; record the person's contribution, its date, the quality criterion, and what it cannot show.

Use a descriptive confidence ladder, not a psychological score for the person. ‘Directly observed’ means a relevant work sample displays the action against a stated criterion. ‘Partially supported’ means an artifact exists but leaves an important condition uncovered. ‘Reported’ means the person's account lacks sufficient corroboration. ‘Unknown’ means evidence is absent or contribution unclear. Do not turn these labels into a decisive four-out-of-five score without a defined rubric. The purpose is to choose the next test. For partial evidence, identify the unproven part of the task. For unknown evidence, design a short sample or ask for a specific behavioral example. Another team member's performance cannot simply be credited to the portfolio owner because they worked together.

Suppose Sara coordinates training and wants to move into learning design. She offers three artifacts: an attendance sheet covering 240 participants, a message explaining poor attendance at one session, and a learning activity created with a colleague. The sheet supports record administration but does not establish learning design. The message may show preliminary diagnosis if it links attendance to timing or access and considers alternatives; saying participants were ‘unmotivated’ without evidence is a leap. The activity is closer to the target role, but Sara's share of the work is unclear. An honest portfolio retains these limitations instead of erasing them or claiming complete readiness.

For each capability claim, create a row containing target task, artifact, individual contribution, quality criterion, descriptive confidence, and gap. In this case, ‘design practice with feedback’ is a target task; the activity is an artifact; Sara's contribution is unresolved; the criterion requires learners to perform a task and receive correction; confidence remains partial until her contribution is documented and the activity is tried. By contrast, ‘coordinate attendance for 240 participants’ is strong administrative evidence but an insufficient learning-design criterion. Administrative skill may still help in the new role, yet transfer requires a link to target action rather than similarity between job titles.

A counterexample tests the judgment. If version history shows the colleague designed the entire activity while Sara only sent invitations, the design claim cannot retain the same confidence; the demonstrated contribution becomes implementation coordination. If an independent review of another activity that Sara designed alone shows clear correction of learner responses, the judgment moves the other way. This is neither a penalty for the person nor a reward for a fluent story. It aligns the claim with the available evidence. Date the portfolio update because a new artifact may change both the gap ranking and the learning route.

Do not collapse three claims: the person performed a task, performed it well, and can transfer that performance to a different context. The attendance sheet may establish that Sara managed a process of a particular size, but record quality still needs a sample check for errors or duplicates. Even excellent record quality does not by itself demonstrate decisions required to design a learning activity: specify an observable objective, choose practice that exposes misunderstanding, and interpret a learner response. It is reasonable to say that disciplined records could help Sara track activity results. It is unreasonable to make record keeping a substitute for design evidence. Transfer is a bridge with evidence on both sides of the roles, not a reassuring phrase on a profile.

Check opportunity bias before judging capability. Sara may lack a solo design sample because work allocation gave her colleague the design role, not because Sara cannot do it. Absence of a sample therefore means ‘unknown,’ not ‘poor.’ Create a fair, bounded opportunity: give her a learning objective, desired behavior, and time constraint; request an activity and answer key; then judge the result by the same rubric used for another candidate. Record unequal access to tools or time if it affects the comparison. A fair test does not lower the quality threshold. It gives comparable opportunities to demonstrate ability before a consequential transition decision.

When sharing the portfolio with a decision maker, state both what the evidence shows and what remains open without stigmatizing the person. ‘Sara independently designed one activity, which has not yet been tried with learners’ is more accurate than ‘expert designer’ and fairer than ‘no experience.’ Name the next review date, who receives feedback, and what action follows if the instructions confuse learners. Revising an instruction may be faster and more valuable than entering a long course. If the gap instead lies in understanding assessment design itself, a deeper learning route may be justified. A useful portfolio is an updatable decision instrument, not a permanent identity judgment.

When two capability artifacts conflict, do not automatically choose the newer one. An older independently reviewed sample on the target task may be stronger than a recent self-description; the reverse can hold if tools or performance conditions changed. Record source, review independence, task relevance, and decision currency for each artifact. A general praise letter does not erase a visible defect in Sara's activity; it raises a question about another successful context that needs a sample. If she later repairs the activity with a version record, retain both versions so learning is visible. This comparison keeps portfolio size from masquerading as evidence strength.