The assumption
If nothing changes, attribution should stay consistent.
- the same campaigns
- the same tracking
- the same reports
So when attribution shifts over time, it feels like something broke.
The reality
Attribution does not remain stable.
Even if you don’t change anything.
Because the system behind it is constantly changing.
What’s actually happening
Attribution depends on:
- how users are identified
- how events are captured
- how sessions are defined
- how data is processed across systems
None of these are static.
They evolve continuously—often without visibility.
As a result, attribution shifts even when your setup appears unchanged.
Where the changes come from
Attribution changes are usually gradual.
They come from small shifts across the system:
- browser and privacy updates — reduce trackable users and shorten data windows
- consent behavior — changes what data is available at all
- platform updates — alter how events and sessions are processed
- site changes — introduce subtle inconsistencies in tracking
- tagging complexity — increases the likelihood of duplication or gaps
Each change is minor.
Together, they reshape how attribution is calculated.
Why this is hard to detect
These changes don’t break attribution all at once.
They create drift.
- numbers still update
- reports still run
- nothing appears obviously wrong
But over time:
- trends become less stable
- discrepancies increase
- explanations become harder
Attribution doesn’t fail—it slowly loses reliability.
Why models don’t fix this
When attribution changes, the instinct is to adjust the model.
But the model is not what changed.
The inputs did.
Switching models may:
- redistribute credit
- smooth out visible differences
But it doesn’t address the underlying shift.
Why platforms diverge over time
Different systems respond to change differently.
- one platform may lose more data due to privacy constraints
- another may adjust how it defines sessions
- another may update attribution windows or logic
Each remains internally consistent.
But they drift apart from each other.
What this leads to
If attribution changes over time, the issue isn’t necessarily a mistake.
It’s the natural result of a system that is evolving without coordination.
Unmanaged systems don’t stay stable. They drift.
What actually stabilizes attribution
Attribution becomes more stable when the system is actively maintained.
That means:
- monitoring tracking quality
- maintaining consistent event definitions
- managing identity and session logic
- adapting to external changes (privacy, browsers, platforms)
Stability is not something you achieve once.
It’s something you maintain.
How to recognize drift
You may be experiencing attribution drift if:
- channel performance shifts without a clear cause
- discrepancies between platforms increase over time
- reports require more explanation than before
- previously stable metrics become volatile
The next step
Before trying to “fix” attribution changes, you need to understand what’s changing underneath.
An Evaluate engagement identifies:
- where attribution is drifting
- what changes are introducing inconsistency
- what is required to stabilize the system
From there, attribution becomes something you can maintain—not just interpret.
Start with Evaluate.
Doug McCaffrey
Designs and maintains analytics systems that remain reliable over time.
Explore how this connects across your data estate: