The paved road test
An internal cloud platform succeeds when teams choose to use it because it is easier than the alternative. It fails when teams use it because they have no choice, and file tickets every time they need something done.
The ticket queue is a symptom. The disease is a platform designed for the platform team, not for the teams it serves.
What self-service actually means
Self-service is not a dashboard. It is a contract: the platform promises that certain operations — provisioning, deployment, scaling, observability — can be completed without waiting for another human.
- Guardrailed defaults: Sensible configurations that work out of the box
- Escape hatches: The ability to customize when defaults are insufficient, without filing a ticket
- Fast feedback: Errors that explain what went wrong and how to fix it
- Documentation that stays current: If the docs are wrong, the platform is broken
Build the platform your worst Tuesday needs
Every team has a bad Tuesday: production is down, a fix is ready, and they need to deploy immediately. The platform should make that Tuesday deployment fast, safe, and auditable. If the bad-Tuesday deployment requires a ticket, an approval chain, and a 30-minute wait, the platform has failed at the moment it mattered most.
A platform team's job is not to build a platform. It is to make every other team faster. The ticket queue is how you know you are not there yet.