The expectation
Most teams expect attribution to:
- match across platforms
- assign clear credit to channels
- provide a single version of truth
When it doesn’t, they assume something is wrong.
The reality
Perfect agreement is not the goal.
Different systems will always:
- observe different parts of the journey
- use different identifiers
- apply different logic
Attribution will never fully align across platforms.
What “good” actually means
Good attribution is not about agreement.
It’s about consistency.
A well-functioning attribution system is:
- internally consistent — the same logic is applied everywhere
- transparent — definitions are clear and understood
- reproducible — results don’t change unexpectedly
- explainable — differences can be understood
What good attribution does
When attribution is working well:
- channel performance trends are stable
- changes in performance can be explained
- decisions are based on direction—not noise
- discrepancies are understood, not debated
You don’t need perfect numbers.
You need reliable signals.
What it doesn’t require
Good attribution does not require:
- perfect tracking
- complete user visibility
- identical numbers across platforms
Those conditions don’t exist in modern analytics.
Where it comes from
Good attribution is not created in reports.
It comes from how the system is defined.
It depends on:
- consistent event structure
- stable identity handling
- aligned logic across systems
- reduced data loss
Why most systems never reach this
Most systems are built for:
- speed
- reporting
- campaign visibility
Not for:
- consistency
- alignment
- long-term reliability
So attribution becomes:
- unstable
- difficult to interpret
- dependent on constant explanation
What this leads to
If attribution is constantly questioned, the issue isn’t the model.
It’s the system behind it.
Good attribution doesn’t eliminate differences.
It makes them understandable.
How to recognize it
You know attribution is working when:
- reports tell the same story over time
- changes in performance can be traced to real causes
- platform differences don’t create confusion
- decisions don’t depend on which report you open
The next step
Before trying to improve attribution outputs, you need to understand how your system is producing them.
An Evaluate engagement identifies:
- where inconsistencies are introduced
- how attribution is being distorted
- what is required to stabilize the system
From there, attribution becomes something you can rely on—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: