The framework

One move, done three ways.

How any part of your team gets better — from level 1 to level 5 — and why all three levers improve the exact same way.

The 1–5 ladder isn’t a grade. It’s one move, repeated.

Most maturity models read like report cards. Runboard treats the scale as the description of a single move, done over and over — getting work out of the one fragile spot and into the team.

Capability · Deploy access Currently level 2
1
2
Target
3
4
5
Ad-hoc
Repeatable
Defined
Measured
Self-improving
Written down in a runbook, but two people still do every release by hand. The move: get it into the pipeline so anyone can ship.
1 Ad-hoc

Stuck in one person. The work lives in one head, one pair of hands, one gut feel. It works by effort and luck.

2 Repeatable

Written down in places, but patchily. Some of it is documented — but it still leans on specific people.

3 Defined

Clear and standard across the team. A new person can follow it without being rescued. The right target for most.

4 Measured

The work reports on itself, and the numbers guide decisions.

5 Self-improving

The system spreads, automates, or tunes itself and gets better without anyone pushing.

One warning before you place anything on the ladder: level 1 assumes the knowledge is at least in someone’s head. Old, stable systems can sit below that — the people who understood them have left, and nothing broke for so long that nobody noticed. Stability is how the gap hides. And beware practices wearing a costume: a rotation that’s staffed and paid but has never resolved anything looks like level 3 and is really level 1 at best. The test for both is cheap: run a drill. Ask “who could fix this today?” If the answer is one name, you’re at level 1. If it’s nobody, you’re at zero — and the clock is already running.

Three kinds of stuck

Three engines, one identical move.

The nine boxes

Where could the effort go?

Cross the three kinds of work with the three levers and you get nine places to spend. Its real value is the habit it forces: work out honestly what kind of problem you have before you spend. The most expensive mistakes are fixing the wrong kind — buying a tool for what’s really a people problem.

Team
builds a role
Tools
builds an automation
Techniques
builds a guardrail
Build
Making new things

Spread build knowledge past the one architect — pairing, design docs, a real second owner.

Automate the path to shipping: build-test-ship pipelines, environments from a script, tests.

Ship without queuing: staged rollouts and one-click undo so releases don’t wait on sign-off.

Run
Keeping what exists working, safe & affordable

No single person holds an outage together: on-call rotations, runbooks, named practising backups.

The system watches itself: monitoring, alerts before customers notice, automatic recovery.

Pre-agreed limits for routine ops: spending caps, change windows, blast-radius rules.

Plan
Connecting the technology to the business

Translation isn’t one person’s job: more than one voice can speak tech to the business.

Make the trajectory visible: dashboards that pull the signals together for the board on their own.

Decide direction without endless meetings: a written decision policy, disagree-and-commit.

Rows: the three kinds of work · Columns: the three levers · Each column builds the same thing every time.

Once it’s scored

The same nine boxes, with a reading.

Run the assessment and every box gets a level (1–5) and a direction of travel — and one box is named as the constraint holding you back. This is the board the Runboard CLI hands back; here it sits at an average of 2.6, with Run · Tools as the thing to raise.

Average 2.6 / 5 Biggest constraint Run · Tools
Find the single spot. Make the work survive without it. Then check that you actually did.