Adoption evidence without telemetry: an opt-in adopters registry we can publish and verify #475

Summary

Shipgate ships no telemetry, deliberately — local-first, static, nothing phones home. The cost of that stance is that we lack a consistent public mechanism for attributable adoption claims: download counts measure CI caches, stars measure sentiment, and the projects that quietly run the gate every day are invisible. Adoption evidence has been this project's #1 named risk in every strategic review; the counting mechanism is therefore a product deliverable, not an afterthought.

Build the lightweight, opt-in, public mechanism:

• ADOPTERS.md (or a pinned discussion): one line per adopter — name/org, repo (or "private"), what they gate — added by the adopter themselves via PR or comment. Being counted is an explicit act by the user; nothing is ever collected.
• An optional README badge adopters can add, linking back to the registry.
• A documented claims policy: any adoption number we publish traces to named, consenting entries in this registry. No entry, no claim.

The precedent is standard OSS practice (adopters files in CNCF projects), and it fits the product's own philosophy: adoption evidence should meet the same bar as capability evidence — explicit, attributable, and verifiable, or absent.

Acceptance

• ADOPTERS.md exists with contribution instructions (what a valid entry contains; private-repo entries allowed with org name only).
• Badge snippet documented.
• Privacy stance stated in the file and README: no automatic collection; public entries are user-initiated and consenting. Private pilot observations are separate, require their own consent, and do not become public adopter entries automatically.
• Linked from the README and the site.
• The registry can include our own verify-self entry, explicitly marked maintainer dogfooding. External adopters are counted separately; a self-entry cannot satisfy an external adoption or repeat-use claim.

What this proves

An entry proves a consenting party made the stated adoption claim. It does not prove continued use, review quality or willingness to pay. Include the entry's date and stated use (local evaluation, advisory CI or blocking CI); do not infer a stronger tier from a badge.

#521 separately measures first value and the next eligible change through the existing design-partner pilot. Names and private observations remain unpublished without explicit consent. #511 is historical-case validation, not either adoption denominator.

Surface discipline

Repository documentation; no code surface, no data collection. Headline metric: attributable public adoption claims, with dogfooding and external entries separated. #521 supplies the workflow observations this registry cannot.

Related

The no-telemetry positioning it protects · the incident-fixture and coverage-report pages (the funnel this registry sits at the end of).