Seeing a payment or booking domain in a source report is a diagnostic lead, not permission to exclude every referral. First establish whether you control and tag both domains, whether navigation preserves the linker and whether the domain is an external service or a genuine referring publisher.
Two official mechanisms
Google’s cross-domain guide describes the same web stream/tag ID and linker-based continuity, including redirect failures. The unwanted-referral guide describes attribution exclusions and payment-return use cases. Both checked 16 October 2026. Suppressing a referral label alone is not proof that identifiers stayed continuous or historical reporting was repaired.
Filled path matrix
Synthetic paths: shop.example and book.example are separate owned sites; pay.example is an external processor; news.example is a genuine publisher. Domain names and outcomes are invented.
On narrow screens, scroll sideways to read the full table.
| Path | Evidence | Decision | Acceptance output |
|---|---|---|---|
| Shop → owned booking | Same stream, consent, linker survives | Cross-domain setup ready to test | Preserved path and redacted debug record |
| Shop → redirect → owned booking | Linker stripped | Repair continuity first | Trace each redirect hop |
| Processor → shop return | External payment context verified | Review unwanted referral rule | Narrow matching condition and test |
| Publisher → shop | Genuine independent referral | Keep the referral | Preserve source attribution |
A controlled test sequence
- Record the exact property, stream and authorised domains. Do not copy identifiers into a public report.
- Use a consented test browser and follow the real link or form into the second owned domain.
- Observe the destination and each redirect for linker preservation; inspect the resulting debug events.
- Test the payment return separately. Keep a genuine referral control to catch an overbroad exclusion.
How to brief the supplier
Ask for one redacted path record per row: entry source, navigation method, controlled domains, stream identity match, consent state, redirect result and observed attribution. Require a separate explanation of unmeasurable traffic. A vendor promising “one person across every device” has gone beyond what this browser-path test establishes.
Interpret the outcome carefully
Passing the owned-domain row supports continuity for that tested browser path only. It does not verify a person, recover denied-consent activity, stitch unrelated devices, reconcile paid orders or prove revenue. The external processor may not permit your tag; do not assume you can configure it as an owned tagged domain.
Boundary and next task
No analytics settings or tracking scripts are changed by this resource. Deployment would need the owner’s relevant access and approval. Once the path is understood, purchase reconciliation compares events with order evidence; it is a separate acceptance decision.
Replay the teaching example
These files run locally without network access. They model only the stated cases, not a live vendor integration. Filled inputs and expected results · Inspectable teaching functions
Save both files in one folder. With Node 22.13 or newer, run:
node decision-check.mjs cross-domain-measurement-choice.json