Rolling out a skills matrix across multiple teams

Phase 1, Pilot (days 1-30)

Pick one team. Not your biggest. Not your hardest. Pick the one whose manager will treat the rollout seriously and will tell you honestly when something isn't working.

In the pilot:

• Adopt the 0-5 scale and write the team-level descriptors (see the competency policy guide).
• Run the 5-day skills audit (see how to run a skills audit).
• Produce the heat-map and the 90-day plan.
• Hold a retrospective. What worked, what didn't, what would you change for team #2?

The pilot is not about getting it perfect. It is about discovering the three things you would have got wrong at scale, and fixing them before they hit 30 teams.

Phase 2, Calibration wave (days 31-60)

Add two more teams. Different from the pilot in some meaningful way, different shift pattern, different site, different function. The goal of this phase is not coverage; it is to discover whether your scoring policy survives translation.

Run all three teams on the same scale, same descriptors, same cadence. After 30 days, hold the first cross-team calibration session.

You will almost certainly find:

• One descriptor needs sharpening because it was written from one team's perspective.
• One skill needs splitting (it was really two skills wearing one label).
• One team's manager rates harder than the others. The descriptor needs work, not the manager.

Update the policy. Reissue. Move on.

Phase 3, Scale wave (days 61-90)

Now you have a policy that has survived three different teams and one calibration. Roll out to the rest of the organisation.

At this scale you need:

• A named matrix owner per team. Usually the team manager; sometimes a 2IC. Not HR. HR sets the policy; teams own the data.
• A monthly calibration cadence across team owners, run by HR or the L&D function.
• A visible dashboard at the leadership team level: only the heat-map summaries, not the individual cells. This shows the rollout is doing something.

At 90 days, you should be able to point to at least one training plan, one cross-training initiative, and one hiring decision that changed because of the matrix. That evidence is what makes the rollout stick.

Four rollout failure modes (and how to dodge them)

1. Secrecy

If staff can't see their own scores, trust collapses and ratings become a paper exercise. Make personal scores visible to the individual on day one. The whole-team view stays inside the team.

2. Direct pay linkage

The moment a score directly drives a pay decision, every rating becomes negotiated, not measured. Capability informs reward (eligibility for a band, for an allowance) but it should not replace performance review.

3. Scope creep

Wave 1 covers 20 skills. By wave 3, somebody has added 60 more and the matrix is unmaintainable. Hold the line. New skills go through a quarterly review, not a Friday-afternoon Slack message.

4. Tool churn

Changing tool mid-rollout (Excel → SaaS → bespoke → back to Excel) is the biggest single trust-killer. Pick the tool that fits your stage (the Excel template for <100 people; PulseAI for larger or multi-site), and commit for at least 12 months.

Communications plan (template)

One paragraph per audience, delivered before the rollout starts:

• To the executive sponsor. "Here is the 90-day plan, the named owner, the budget, and the 3 decisions this matrix will change at our next board cycle."
• To line managers. "Here is what we are asking you to do, what we are not asking, and the time commitment."
• To everyone in scope. "Here is the scale, here is what your ratings will and will not be used for, here is who to ask if you have concerns."

How to measure a rollout

• Coverage: % of in-scope people with at least one current rating (target: 95% by day 90).
• Freshness: % of cells refreshed within their cadence (target: 90%).
• Decisions: count of training, mentoring, or hiring decisions visibly informed by the matrix (target: at least three per team).
• Trust: a one-question pulse, "do you believe your matrix rating reflects you fairly?" (target: 75%+ agree).