The Default Response
When attribution doesn’t make sense, most agencies respond the same way:
- switch models
- compare platforms
- build new reports
They ask:
“Which attribution model should we use?”
It feels like the right question.
It isn’t.
Attribution is not a reporting feature
Attribution isn’t something you choose.
It’s something that emerges from how your system is defined.
It reflects:
- how data is collected
- how sessions are defined
- how users are identified
- how events are structured
Long before any model is applied.
For a broader view of how attribution is actually determined:
Where It Breaks
Agencies operate inside tools like:
- Google Analytics 4
- ad platforms
- reporting dashboards
Within them, they accept:
- default definitions
- black-box logic
- fragmented identity
Then attempt to “fix” attribution in reporting.
Why This Fails
Attribution is not a reporting feature.
It is the outcome of upstream decisions.
If those decisions are inconsistent:
- models won’t align
- platforms will disagree
- reports will contradict
No model can resolve that.
The Illusion of Model Choice
First-click
Last-click
Data-driven
Position-based
These feel like strategic choices.
But when the inputs are inconsistent:
you’re applying different interpretations to unstable data
The output changes.
The problem doesn’t.
Where attribution is actually determined
Attribution becomes meaningful only when the system is defined.
It depends on:
- how identity is handled
- how sessions are structured
- how events are defined
- how logic is applied across systems
Where agencies go wrong
They optimize for:
- speed
- outputs
- platform alignment
Instead of:
- consistency
- structure
- definitional clarity
The result:
- dashboards that look correct
- numbers that don’t agree
The actual failure point
Attribution breaks when:
- identity is fragmented
- sessions are inconsistent
- events are loosely defined
- channels are unstructured
At that point:
attribution isn’t inaccurate
it’s undefined
Why platform comparisons fail
Agencies try to reconcile:
- Google Ads
- Google Analytics
- other tools
But each platform:
- defines users differently
- reconstructs journeys differently
- applies its own logic
So:
attribution becomes interpretation layered on inconsistency
What good attribution looks like
Not:
- perfect agreement across platforms
But:
- consistent logic within your system
- transparent definitions
- reproducible results
When the system is stable:
- attribution becomes explainable
- differences become understandable
- decisions become more reliable
What this leads to
If attribution doesn’t make sense, the issue isn’t the model.
It’s the system behind it.
Optimizing attribution without addressing the system only makes the problem harder to see.
Final thought
Most agencies believe:
attribution is about choosing the right model
In reality:
attribution reflects how well your system is defined
If this feels familiar
If attribution:
- changes depending on the report
- requires constant explanation
- leads to debate instead of decisions
You don’t have a model problem.
You have a:
system definition problem
The next step
Before changing models or comparing platforms, you need to understand how your system is actually behaving.
An Evaluate engagement identifies:
- where attribution is being distorted
- how inconsistencies are introduced
- what is required to improve reliability
Start with Evaluate.
Doug McCaffrey
Designs and maintains analytics systems that remain reliable over time.
Explore how this connects across your data estate: