Archive note, April 2026: This post is based on Model Hub measurement-boundary notes and BDC analytics documentation.
Analytics dashboards are very good at encouraging overconfidence.
A chart shows visits. A campaign shows clicks. A tracked link shows engagement. It is tempting to keep walking the conclusion forward until the dashboard appears to prove revenue, appointments, leads, or sales.
Sometimes it does. Often it does not.
The measurement notes for Matomo and tracked assets were written to keep that boundary honest.
What Was Actually Measurable
The system could directly measure controlled signals:
- Tracked asset visits.
- Pageviews.
- CTA clicks through controlled redirects.
- Campaign identifiers.
- Source and channel metadata when present.
- Bounce rate and session behavior on controlled pages.
Those are useful signals. They can show whether an asset was opened, clicked, or engaged with. They can compare campaigns and channels. They can help improve creative decisions.
They do not automatically prove downstream business outcomes.
The Missing Loop
The hard constraint was simple: no full access to client CRM systems and no direct control over all destination websites.
Without that loop, the system could not honestly claim that a specific email asset caused a sale or appointment. It could show that a user clicked. It could show that traffic arrived. It could sometimes show engagement after arrival. But the final business result often lived outside the measurement boundary.
That boundary should be visible in the reporting.
Four Tiers
The measurement model used four tiers.
Directly measurable signals are observed by the system itself. Clicks and controlled pageviews belong here.
Directionally inferable signals suggest performance but do not prove outcomes. A campaign may produce stronger destination engagement than another, but that is still a proxy.
Partner-reported outcomes are real only if the partner reports them. They may include leads, calls, appointments, showroom visits, or sales, but they should be labeled as externally reported.
Not currently observable outcomes are the ones the stack cannot support yet. Pretending otherwise is bad analysis.
Language Matters
The reporting language became part of the product design.
Good language:
- This asset generated tracked visits.
- This channel produced stronger destination engagement.
- Client-reported lead lift coincided with the campaign.
- This campaign appears directionally stronger.
Bad language:
- This asset caused sales.
- Matomo proves revenue.
- This CTA definitively produced appointments.
That may sound like semantics. It is not. It is the difference between credible measurement and sales theater.
The Strategic Value
Incomplete attribution is still useful.
You can compare asset families. You can identify weak calls to action. You can see which campaigns earn clicks. You can decide which creative patterns deserve more work. You can build reporting that is ready to absorb better data later.
The trick is not pretending the current system sees more than it sees.
Honest analytics are less exciting than inflated analytics. They are also more useful.