Aren’t ADRs just paperwork that slow teams down?

Bad ADRs are. Good ADRs are lightweight, live next to code, and tie directly to templates and pipelines. They reduce churn by preventing re‑litigation of settled decisions and accelerate changes by clarifying default paths. Keep them short, scoped, and linked to paved roads.

How do we handle teams that need to deviate for good reasons?

Create an exception ADR with an owner, clear scope, and an exit plan. Deviations are fine, but the team owns oncall and compliance. Revisit quarterly. Many “exceptions” end up back on the paved road once the experiment’s value is proven—or not.

What if we’re already deep in drift—where do we start?

Pick one high‑leverage baseline (e.g., `node-service` or `terraform-module`). Ship a great template, enforce a minimum set of guardrails, and migrate the top 20% of services by change volume. Wins there pay for the rest. Don’t try to standardize everything at once.

Will paved roads stifle innovation?

They stifle novelty in the undifferentiated heavy lifting: logging, health checks, deploys, infra. That’s intentional. They free up cycles to innovate where it matters—product features and ML models—not in writing a new `Dockerfile` for the 15th time.

Which tools do you recommend to operationalize this?

Backstage for discoverability and scaffolding; GitHub Actions reusable workflows; ArgoCD for GitOps; Gatekeeper or Kyverno for policy; Renovate for upgrades; Sourcegraph/OpenRewrite/jscodeshift for refactors; Terraform with a module registry for IaC. Swap equivalents as needed; the pattern matters more than the brand.

Platform-productivity · Oct 1, 2025 · 9 minute read

Stop Drift: ADRs and Paved Roads That Make Safe Refactors Boring

Capture decisions once, codify them into golden paths, and make the default the safe path. ADRs set the rules; paved roads automate them.

Back to all posts

Stop Drift: ADRs and Paved Roads That Make Safe Refactors Boring

Key takeaways

Implementation checklist