Most internal developer platforms fail not on technology but on adoption. This team shipped a working IDP in 120 days and won 18 of 22 teams without a single mandate.
The wrong status quo: build the IDP like an infrastructure program and ship it like a decree, so it sits "almost ready" for nine months while teams keep filing tickets and copying YAML out of Slack.
The better approach: make the platform a product, win the next team before adding the next feature, and let voluntary adoption be the scoreboard.
Team Topologies defines the relationship between a platform team and the teams it serves as a product relationship, not a help-desk or gatekeeper one.
A golden path is a pre-built, opinionated, validated workflow for a common task like "deploy a new microservice" or "add a new database.
Voluntary adoption is the only kind that lasts. Make the platform fast and reliable enough that engineers pick it because it makes their day easier.
Pick the single most common workflow. Map every step engineers do today: every ticket, every approval, every copy-pasted config, every wait.
Roll out to more teams one at a time, with paired support from the platform team for each team's first week.
Add the second golden path. Provisioning, secret management, and runbook automation are the usual choices.
With two proven paths, widen the rollout and get the platform team out of the critical path.
If your IDP has been "almost ready" for nine months, the answer is not more features.
A first golden path usually takes 4 to 6 engineers. Scaling adoption across the org adds a product manager and 2 more engineers. We have shipped MVPs with smaller teams when they are paired with an embedded engagement.
Both. Buy for capabilities with mature commercial options like CI/CD platforms and secret management. Build the opinionated workflow that is unique to your engineering culture. That is where a platform earns its keep, and where off-the-shelf tools cannot match a paved road designed around how your teams actually work.
Watch the first-week paired sessions. The friction engineers hit the first time they really use a path is your leading indicator. Adoption is the lagging one. If the first week is smooth, the numbers follow.
We re-scope to a single golden path, sunset the unused capabilities, and rebuild adoption from a narrow base. We have run this rescue four times in 2025 to 2026. The root cause is almost always the golden cage, where the platform replaced one wait-on-someone workflow with another, and the fix is less surface area, not more.
Spotify's Backstage is the reference. It cut developer cognitive load by about 40%, supports more than 14,000 services, and lets a new engineer stand up a production-ready service in under 10 minutes, work that used to take more than two weeks.
