Backstage stopped being "just a UI" sometime last quarter. The big, under-remarked shift is that catalogs, templates, and scorecards are now being positioned as the canonical control plane for a multi-layer internal developer platform (IDP) stack — and teams are wiring DORA metrics and Four Keys telemetry directly into those surfaces to run the platform like a product.
Say it plainly: platform teams are no longer focused on stove-piped plugins and bespoke scaffolding. They're building opinionated, measurable products that sit between raw infrastructure and developer-facing services. The research and recent guidance circulating in the community separates Infrastructure Platform Engineering (IPE) — the discipline that turns clusters, networks, and storage into a consumable product — from higher-level DevEx platform work. That's an important demarcation because it forces platform teams to choose scope: own the infra primitives or own the developer journey. Both have engineering cost, but only one scales unless you standardize it.
The practical fallout is straightforward. Backstage's Software Catalog and Scaffolder plugin become the wiring — service metadata, ownership, templates, and automated scorecards — while a separate IPE layer exposes opinionated primitives (cluster profiles, environment provisioning, default security posture). Teams then glue DORA metrics collection and the open-source Four Keys project (from Google Cloud) into that catalog so every golden-path template emits consistent telemetry: deployment frequency, lead time for changes, change failure rate, and time to restore. Four Keys is a common integration point for IDPs that want to prove impact instead of producing screenshots.
This isn't academic. If your Scaffolder template creates a CI pipeline, it should also create the observability bits, the SLO/alerting bits, and the telemetry hooks that feed DORA metrics. When the template enforces those defaults, your platform can measure whether using the golden path actually moves the needle on lead time or production stability. DORA/Accelerate research and related industry guidance back this up: platforms that bake telemetry into developer workflows see measurable improvements in deployment frequency and lead time. Link that to product management priorities and you suddenly have a data-driven roadmap for platform investment.
Two more shifts to note.
First: data and AI workloads are now first-class in IDPs. Platform teams are being asked to provide self-serve feature stores, model deployment pipelines, and governed data access through the same catalog and template tooling. That means your Backstage templates will need to create DB migrations, dataset registrations, model CI pipelines, and model governance metadata — not just a microservice stub. Treating ML as a special case is a losing bet.
Second: security and observability are embedded features, not bolt-ons. Platform teams must own standardized security baselines, automated controls (policy-as-code), and unified logging/metrics/traces surfaces that templates expose. If your golden path omits an automated control or a standard policy, the platform isn't doing its job — it's outsourcing the problem to developers.
Here's an opinion you can take to the platform steering committee: stop measuring platform success by tickets closed or tools deployed. Measure platform adoption by DORA-style outcomes tied to golden-path usage. If your platform product roadmap doesn't show how a template improves lead time or reduces change failure rate, it's a feature catalogue, not a platform product. Teams that treat Backstage as a plugin playground instead of the catalog + control plane for observable, governed templates will be the ones rebuilding governance in six months.
The sensible next move is obvious: treat the IDP as a product with clear KPIs, instrument every template with DORA/Four Keys hooks, and expand catalogs to include data and ML artifacts. This is overdue. The platform world finally has the telemetry and product discipline to stop arguing about toolchains and start owning developer outcomes — which, frankly, is the whole point.
If you run a platform team: pick two golden paths, add DORA hooks, and make the data visible to engineering managers within 90 days. If you don't, someone else will ship a packaged IDP that does, and your hard-won platform influence will evaporate.
Sources
- Platform Engineering Blog – disciplines overview (Infrastructure, DevEx, Data, Security, AI, Leadership, Product Management, Observability)
- Google Cloud – New platform engineering research report (platform engineering as a strategic trend, relation to DevOps and DORA)
- DORA Four Keys GitHub repository (open-source implementation of DORA metrics, commonly integrated into IDPs)