IndieWeb field note ·

Clean-link attribution for independent sites.

A distribution experiment can be measured without adding campaign parameters to the link. Match the actual referring host with a dedicated landing path, store only an aggregate-safe visitor identifier, and label the result honestly when the referrer disappears.

By Traction Garden

The pattern

Two ordinary request facts form the experiment key.

  1. Publish a distinct, useful landing page for one distribution experiment.
  2. Share its canonical URL without campaign or experiment parameters.
  3. When a human pageview arrives, compare the normalized referrer host and exact landing path against a small server-side mapping.
  4. Assign the strategy only when both values match. Otherwise keep the visit unattributed.
  5. Count distinct visitors with a secret-keyed hash; discard the raw address before storage.
Minimal implementation

Prefer a narrow allowlist over broad inference.

function inferExperiment(path, referrerHost) {
  if (
    path === "/research/clean-link-attribution" &&
    referrerHost === "news.indieweb.org"
  ) return "indienews-clean-link-v1";

  return null;
}

The mapping is intentionally specific. A visit to another page from the same host is not this experiment; a visit to this page without that host is not assigned to IndieNews.

Trade-offs

Clean links exchange attribution coverage for restraint.

Tracked queryexample.net/post?utm_source=communitySurvives missing referrers, but the identifier travels in copies, screenshots, logs, and bookmarks.
Referrer + landingexample.net/postKeeps the shared URL clean, but loses attribution when a browser, app, or privacy policy strips the referrer.
Landing onlyexample.net/postCan prove the asset was visited, but cannot distinguish the catalog from search, direct sharing, or another link.
What not to claim

A missing referrer is not evidence of “direct” intent.

Browsers, native apps, redirectors, and referrer policies can suppress the source. Record the visit as unattributed instead of guessing. Likewise, a catalog accepting a Webmention proves delivery to the catalog—not that a person saw or clicked it.

Live implementation

The evidence stays inspectable.

Traction Garden uses this exact pattern for this page. The public scorecard separates upstream publication evidence from downstream distinct visitors. It stores HMAC-derived visitor hashes, never raw IP addresses, and excludes known bots and internal verification.

Syndication

This post is submitted by Webmention.

Also posted on IndieNews. The submission is accountless and standards-based: IndieNews reads this page's h-entry, then returns a public catalog permalink.