I had already accepted that attribution was incomplete. That still left a practical problem: what language should appear next to a number in a report?
I settled on four labels. They are not sophisticated, but they stop a dashboard from quietly upgrading evidence as it moves from an event log to a presentation.
Measured
Measured means the system observed the event on infrastructure we control.
A redirect recorded a click. Matomo recorded a landing-page visit. A campaign identifier arrived with the request. These are still imperfect observations—a scanner can follow a link and a tracking pixel can load without anyone reading—but the event itself exists in our data.
I report the event, not the imagined person behind it.
Inferred
Inferred is for interpretation built from measured events.
If one asset family repeatedly produces longer destination sessions, I can call it directionally stronger. If traffic rises around a send, I can describe the timing. I cannot turn either pattern into a lead or sale without another source of evidence.
Inference is useful. The label keeps it from impersonating measurement.
Partner-reported
Clients sometimes know what happened after a visitor leaves systems I control. They may report calls, appointments, or sales during a campaign period.
Those outcomes belong in the report with their origin attached: partner-reported. The label does not diminish the result. It prevents readers from assuming Matomo independently observed it.
Not observable
Some questions do not have an answer in the current stack.
When a tracked click lands on a dealer site I cannot instrument and there is no CRM integration, the exact path from asset to purchase is not observable. A blank in that part of the journey is not zero, and it is not permission to estimate.
How It Reads In Practice
The report can now say:
642 destination visits measured
stronger session depth than the prior send inferred
client noted increased appointment activity partner-reported
campaign-attributed vehicle sales not observableThat small vocabulary is more useful to me than another elaborate attribution diagram. It tells the reader where the evidence came from and where it stops. If better integrations arrive later, an outcome can move categories without rewriting the history of what we knew at the time.