Writing a 0-5 competency policy

The 0-5 scale (canonical descriptors)

Use these as your starting descriptors. Edit the noun phrases to match your context (replace "the standard" with the specific SOP / regulation / framework your team works to).

Evidence rules (what unlocks each level)

Without an evidence rule, ratings drift upwards. Pin them with this minimum:

• Level 0-2: stated by the individual; agreed by the line manager. No further evidence required.
• Level 3: evidence of independent delivery to standard within the last 12 months. Examples: completed jobs, signed-off work, audit pass, exam.
• Level 4: evidence of handling at least one non-standard case in the last 12 months, plus evidence of teaching the skill to another person.
• Level 5: evidence of setting or revising the standard (procedure, SOP, training material) and of certifying others against it.

The point of evidence rules is not bureaucracy. It is to make sure a "4" awarded in one corner of the organisation is the same shape as a "4" in another.

Who rates whom?

The three-rater pattern is the most reliable:

1. Self-rating first. Honest, recorded, dated.
2. Line manager second, without seeing the self-rating.
3. Subject-matter expert / peer third, on skills the line manager doesn't personally hold at the required level.

When the three ratings agree, the score sticks. When they differ by more than one level, a 15-minute conversation reconciles them. Most reconciliations land between the self-rating and the manager rating; the SME is the tiebreaker on technical skills.

Refresh cadence and stale ratings

Set a review-by date on every cell. The matrix should automatically flag any rating older than the cadence (in our Excel template and PulseAI, this is a configurable amber/red flag).

Recommended cadences by skill type:

• Regulatory / safety, every 6 months (or per the regulation's own cadence).
• Technical / tool-based, every 12 months.
• Leadership / soft, annual, aligned to the performance review cycle.

Calibration sessions

A calibration session is a 30-minute meeting between managers who rate against the same scale. Pick three skills, look at the score distributions side by side, and ask: "do these distributions feel right relative to each other?" If they don't, the descriptor needs sharpening, not the scores.

Run calibration quarterly for the first year, then twice a year. Without it, scales drift and the matrix gradually loses trust.

Worked example, three skill descriptors in three domains

Skill: CNC machining

• 3, Independent. Sets up, runs, and edits programs for the standard part family without supervision. First-off inspection passes >95% of the time.
• 4, Proficient. Sets up and runs unfamiliar part geometries; recovers from tool break or program fault without escalation.
• 5, Expert. Writes the team's machining standards; trains apprentices; certifies new-starter checklists.

Skill: Customer escalation handling

• 3, Independent. Owns a complex customer escalation through to resolution without manager intervention.
• 4, Proficient. De-escalates the high-value or media-sensitive case; documents lessons learned.
• 5, Expert. Coaches the team on escalation patterns; revises the playbook quarterly.

Skill: IFRS lease accounting

• 3, Independent. Prepares journal entries and disclosures for standard lease portfolios to audit-ready standard.
• 4, Proficient. Handles unusual lease modifications, sale-and-leaseback, and discount-rate changes without external advice.
• 5, Expert. Owns the team's IFRS-16 policy paper; trains the wider finance function; advises external auditors.

Governance one-pager

Keep the policy short. One page is the maximum that anyone will actually read. Include only:

1. The scale and descriptors.
2. The evidence rules per level.
3. The who-rates-whom flow.
4. The refresh cadence.
5. The calibration cadence and owner.
6. Who owns the matrix (named person + deputy).

Everything else is a footnote. The point is for every rater to be able to read the page in 4 minutes and start scoring.