Skip to main content

Radar contribution & review

How blips get onto the tech radar and how they move between rings. The radar is data in Git (radar.json), so this is a Git process: propose in a PR, review on a cadence, record consequential moves as ADRs. The radar and the security allowlist are the same source of truth — a blip's ring is also its permission to be in the stack (enforced by the intake gate).

Proposing a blip​

A blip proposal is a PR against radar.json whose description answers four things:

  • Problem — what it solves that the current adopt ring doesn't. "It's newer" is not a problem statement.
  • Evidence from real use — ours, not a vendor's: a spike result, a squad's trial notes, a harness run. A blog post is a reason to assess, never to trial or adopt.
  • Ring recommendation — adopt / trial / assess / hold, with the reasoning.
  • Migration note — what it replaces or overlaps with, and what happens to the incumbent. Every addition without a subtraction grows the estate we just spent STOP shrinking.

Anyone in any squad, UK or India, may propose. New third-party components in a proposal still pass the intake gate checks — the radar PR and the intake approval are usually the same PR.

Review cadence​

  • Quarterly cross-UK/India review — the Toolkit Guild walks the whole radar with squad reps from both sites: pending proposals decided, stale assess blips expired, hold items checked against the remediation backlog. One session, both time zones represented — the radar must not fork by geography.
  • Between quarters — uncontroversial moves (a new assess entry, an assess → trial with evidence) are approved asynchronously by the Guild in the PR. The quarterly is the sweep, not the queue.

The fast lane for AI tooling​

The AI models, agents & tools quadrant moves monthly, not quarterly — a quarterly-only cadence would leave us permanently a model-generation behind. AI-tooling blips are reviewed within the monthly AI-governance review, and their evidence bar is specific: a golden-task harness run per the tooling-evaluation playbook, not sentiment. The fast lane changes the clock speed, never the evidence standard.

Ring moves with consequences​

  • Any move affecting a live product — a framework to hold, a platform out of adopt — is recorded as an ADR linked from the blip, naming affected products and the migration expectation. The blip says what changed; the ADR preserves why.
  • Hold is a verdict, not a shelf — placing a blip on hold creates or updates entries in the EOL inventory / remediation backlog for products still using it.
  • No silent moves — every ring change is a PR with a reviewer; the radar's Git history is its audit trail.

Standards referenced: ThoughtWorks Tech Radar (technique), ADRs.