All insights

Platform engineering

What an internal developer platform should give a team on day one

An internal developer platform is successful when a developer can take a new service to production without opening a ticket. Everything else is detail.

The golden path

The core of a platform is one well-paved route from idea to running service. It should not be the only route, but it should be the easiest one, so that teams choose it because it saves them time.

Five things to have on day one

  1. A service template. A new repository arrives with a build, tests, a container image and a deployment definition already wired together.
  2. Environments on request. A pull request or a single command creates a working environment, with no manual cluster access.
  3. GitOps delivery. Git is the source of truth for what runs. A reconciler such as Argo CD or Flux applies changes and corrects drift, so every deployment has an author and a history.
  4. Observability by default. Metrics, logs and a starter dashboard exist the moment a service is deployed, along with sensible alerts.
  5. Documentation that matches reality. A short guide and a catalogue of what is running, kept next to the code so it stays current.

Keep it thin

The common failure is building too much. Start with the smallest platform that removes the most waiting, then extend it based on what teams ask for. Treat the developers as customers: talk to them, watch where they get stuck and fix that first.

Measure one thing

Track the time from creating a repository to the first production deployment. If that number falls from weeks to hours, the platform is working. If it does not move, more features will not help.