Spotify's squad-based engineering culture, with hundreds of small autonomous teams each owning their own services, produces a natural side effect: a lot of services, built and operated independently, with no single place to find out what exists, who owns it, or whether it's healthy. By the mid-2010s, Spotify's internal engineering surface had grown into thousands of microservices, and simply answering "who owns this thing and is it okay" had become its own recurring source of friction, especially for anyone crossing team boundaries — onboarding a new engineer, tracking down an on-call owner, or auditing which services still used a deprecated library.
The discoverability problem behind fragmentation
The underlying issue wasn't that Spotify lacked tooling — it was that tooling was scattered across many disconnected systems: one place for CI status, another for service ownership, another for documentation, another for infrastructure provisioning. An engineer trying to understand a service they didn't own had to hop between multiple internal tools, each with its own login and its own partial view, to piece together a full picture. That friction discouraged cross-team contribution and made platform teams' job of enforcing standards — consistent monitoring, up-to-date dependencies, documented ownership — much harder, because there was no central inventory to check compliance against.
A single pane of glass, built around a software catalog
Spotify's answer, Backstage, centered on a software catalog: a structured registry of every service, library, and infrastructure component, each with declared metadata like its owning team, its documentation, its dependencies, and links to its CI/CD status. On top of the catalog, Backstage added software templates for scaffolding new services with sane defaults baked in, and a plugin architecture letting teams surface their own tools — CI dashboards, cost data, incident history — inside the same portal rather than as yet another standalone destination. The catalog turned "who owns this and is it healthy" from a multi-tool investigation into a single lookup.
Open sourcing it, and why platform teams mattered
Spotify open sourced Backstage in 2020 and donated it to the Cloud Native Computing Foundation in 2022, and it's since become one of the more widely adopted developer portal frameworks industry-wide. The broader lesson Spotify drew, echoed in its public engineering talks, is that platform teams earn their keep by removing this kind of cross-cutting friction for everyone else — a catalog and a portal don't ship product features directly, but they reduce the tax every team pays navigating the company's own infrastructure, which compounds across hundreds of squads.
Operational gotchas of a developer portal
Backstage works when the software catalog is true: owners, on-call, repo, and deploy target for each service. It fails when it is a wiki with YAML that nobody updates, and then the portal becomes the place where information goes to die. Mid-size steal: a catalog that is generated from running systems and CODEOWNERS, with a weekly stale-owner report, before you install a plugin marketplace.
The concrete failure mode is making Backstage a gate for deploys on day one, so product teams invent shadow pipelines. Another is 40 plugins and no paved path; cognitive load exceeds the old wiki. Operational gotcha: authz. A portal that can trigger jobs is production access. Treat it like an IDP. Spotify could staff a platform experience team. You should pick five golden paths — new service, new data pipeline, on-call signup, docs, CI — and finish them. Scorecards that shame teams without fixing CI minutes will be ignored or gamed. Steal the idea of a homepage for engineering, not a second Jira. If your company has 12 services, a well-kept README index plus service.yaml in each repo may be enough until discoverability actually hurts. When you do adopt Backstage, the catalog freshness SLO is the only SLO that matters; everything else is a theme.
What you can borrow
- As the number of independently owned services grows, ownership and discoverability become their own engineering problem — don't wait for onboarding pain or incident response confusion to force the issue.
- A structured catalog with clear metadata (owner, docs, dependency links) is worth building before you need it for compliance audits or incident response, not after.
- Consolidating scattered internal tools behind one portal reduces context-switching cost for engineers working outside their own team's usual territory — that cost is easy to underestimate because it's paid in small, frequent increments.
- Platform investment that doesn't ship product features can still have a strong ROI if it removes friction that every team pays repeatedly.