Multi-channel attribution assigns credit across touchpoints in a B2B journey. No model is accurate, because much of the journey is unobservable, so the practical use is comparing directional performance consistently rather than establishing true contribution.
Attribution debates consume enormous energy in B2B marketing teams and rarely resolve, because they are arguments about a question with no correct answer.
Meaningful parts of a B2B journey are invisible: a recommendation in a private conversation, a mention in a community, an AI assistant summarizing your category. No model captures any of it.
Why no attribution model is correct
Every model encodes an assumption about which touchpoints matter. First-touch credits discovery, last-touch credits closing, linear treats everything as equal, and none of these is true.
Add the unobservable touchpoints and precision becomes impossible in principle rather than in practice. Accepting that early prevents a great deal of wasted argument.
What attribution is actually good for
Comparison under a consistent method. If the same model is applied over time, changes in a channel's measured contribution are informative even though the absolute number is wrong.
That is enough for the decision you usually face, which is where to move the next increment of budget rather than what each channel deserves in some final accounting.
Which model should you use?
- First-touch. Useful when the question is what creates awareness in a market that does not know you.
- Last-touch. Simple and heavily biased toward capture channels that intercept demand created elsewhere.
- Multi-touch. More balanced, more complex, and only as good as your tracking of the touchpoints you can see.
- Holdout testing. The only method that measures actual incrementality. Slower, harder, and considerably more truthful.
Get the plumbing right first
Attribution depends on joining anonymous behavior to a known contact to a CRM opportunity. If that chain breaks, the model is computing over incomplete data and no sophistication rescues it.
Consistent UTM conventions, an identifier passed through to the CRM, and filtered internal traffic matter more than model selection. Most attribution problems are data problems.
Use it alongside asking people
A self-reported question on your form, asking how they heard about you, captures things no tracking will. The answers routinely name channels attribution reports as negligible.
Treat that as a corrective. When self-reported and modeled attribution disagree sharply, the gap usually locates a channel doing invisible work.
- ✓ No attribution model is correct. Much of a B2B journey is unobservable in principle.
- ✓ Use one model consistently to compare over time, not to establish true credit.
- ✓ Most attribution problems are data problems. Fix the join before choosing a model.
- ✓ Self-reported attribution catches what tracking cannot. Disagreement locates invisible work.
FAQ
Which attribution model is best for B2B SaaS?+
None is accurate. Pick one appropriate to your question, apply it consistently, and use holdout tests when a decision is large enough to justify measuring real incrementality.
Why does attribution never match what sales says?+
Because significant parts of a B2B journey are invisible to tracking: private recommendations, community mentions, and AI-generated summaries. Sales hears about these, and no model records them.
Should you ask customers how they heard about you?+
Yes. Self-reported attribution captures channels tracking cannot see, and sharp disagreement with your model usually identifies a channel doing invisible work.