What is platform engineering and what problem does it solve?
Platform engineering: Gartner predicts 80% of large software companies will have platform engineering teams by 2026. Problem solved: as organizations adopt microservices and cloud-native, each product team must be a full-stack expert (Kubernetes, Terraform, observability, CI/CD, security). Cognitive overload slows delivery. Platform team builds the IDP (Internal Developer Platform) — the "paved road" — so product teams get opinionated, working defaults for all infrastructure concerns. Key principle: treat developers as customers. Platform team measures success by developer productivity, not infrastructure uptime.
2 / 5
What is a golden path in platform engineering?
Golden path: the "happy path" that just works for most teams. Inspired by Spotify's "paved road" concept. Elements: Service template: cookiecutter or backstage scaffold creates a new repo with: Dockerfile, CI pipeline (GitHub Actions), Helm chart, observability (Prometheus annotations, OpenTelemetry SDK), Dependabot. Self-service deployment: developer opens a PR; CI builds, tests, pushes image, deploys to dev. No ops ticket required. Sensible defaults: pod resource limits, liveness probes, HPA, PodDisruptionBudget, network policy — all pre-configured. Teams can diverge from the golden path but must own the maintenance burden of their custom setup.
3 / 5
What is Backstage and what does it provide to platform teams?
Backstage (Spotify, open-sourced 2020, CNCF): three main features: Software Catalog: all services, libraries, APIs, data pipelines in one searchable registry. Each entity has: owner, dependencies, docs link, API spec, deployment status. Solves "what services do we have and who owns them?". Software Templates (Scaffolder): self-service new service creation. Developer fills a form, Backstage creates the GitHub repo from the golden path template, registers in catalog. TechDocs: docs-as-code (Markdown in repo) rendered in Backstage. Plugins: 200+ community plugins — PagerDuty, Kubernetes, ArgoCD, Grafana, SonarQube. Platform team configures Backstage; product teams use it as their engineering portal.
4 / 5
What does developer self-service mean in platform engineering?
Developer self-service: the goal of platform engineering. Measured by: time to first deployment for a new service (target: <1 day), unblocked work (no ops tickets required for standard operations), cognitive load (developer doesn't need to know Kubernetes to deploy). Self-service capabilities: Environments: create preview/staging environment from a PR — automatic teardown on merge. Databases: provision a dev PostgreSQL instance in minutes via platform UI. Secrets: request secrets from Vault through automated workflow, injected at runtime. Feature flags: enable/disable via platform UI without deployment. Observability: dashboards auto-created when a new service is registered in the catalog. Team topologies: Platform team = "Enabling team" (reduces cognitive load for Stream-aligned teams).
5 / 5
How do platform engineering teams measure success?
Platform engineering success metrics: DORA metrics: platform should improve all four — faster deploys (tooling), lower lead time (self-service), faster MTTR (better observability), lower change failure rate (automated testing/rollback). Developer NPS: would you recommend the platform to a colleague? Treat developers as customers. Time to first production deploy: new hire / new service should reach production within 1 day using golden path. Adoption rate: what % of teams use the golden path vs custom setups? Support ticket volume: platform-related tickets should decrease as self-service improves. Toil reduction: hours engineers spend on repetitive platform tasks. Platform team anti-pattern: measuring infra uptime (ops mindset) instead of developer productivity (platform mindset).