Multi-site attribution works when every site accepts the same bounded source vocabulary, preserves it through intake, and reports later commercial states against the original site and source without inventing cross-context identity.
Use one source vocabulary across the portfolio
Separate sites become impossible to compare when each one names sources differently. The attribution contract defines accepted values for source, medium, campaign, content, referring host, landing route, and site slug. Unknown or stripped values remain unknown instead of being reassigned to a preferred channel. This makes a portfolio total reproducible from the site-level records.
Plausible supports common UTM parameters and uses them in campaign attribution while removing them from the stored page path. The implementation can follow that separation: acquisition fields describe how the visit arrived, while route fields describe what the visitor read. A later form state retains the bounded source fields without copying the full referring URL or private submission text.
Keep site identity stable through intake
The public form emits a submission identifier and site slug to the same first-party collector used for page events. The intake backend records the identifier beside the new lead record, then emits later states such as qualified, unqualified, proposal requested, ordered, or closed. Those backend events make it possible to compare commercial outcomes while the private content stays outside the analytics stream.
Two different hostnames may create two separate sessions for the same browser, and that can be the correct privacy boundary. Plausible documents that hostname participates in unique-visitor recognition and that events from two hostnames produce separate sessions. A portfolio report can aggregate site and source totals without claiming that the sessions belong to one person.
Report attribution limits beside the totals
Direct visits, privacy controls, copied links, redirects, and long delays can remove or change acquisition context. The weekly report states how many later states retained an original source, how many remained direct or unknown, and which joins failed. Source attribution then supports comparison within its observed coverage instead of assigning certainty to every inquiry or order.
Where the service stops
Reality Contact, LLC implements and verifies the approved measurement system, but does not supply legal advice, choose consent on the buyer's behalf, identify visitors across unrelated contexts, contact leads, or decide which offer survives. The buyer approves event definitions, privacy and retention rules, exclusions, qualification states, and uses the first cohort report to keep, revise, hold, or retire each offer. This is analytics implementation and operational reporting, and it does not replace privacy counsel, consent review, security review, or the buyer's own commercial judgment. Reports describe recorded events and their known coverage; they do not establish why a visitor acted or whether an unobserved person had commercial intent.
Sources: Plausible campaign parameter documentation; Plausible Events API hostname behavior.