B2B measurement

B2B marketing attribution, and where it breaks

Attribution tries to give credit for a B2B deal to one or more touches. First-touch, last-touch, and multi-touch models all fail in different ways because committees, dark social, and long cycles do not look like a retail checkout.

First touch and last touch are simple, and biased

First-touch gives the origin story: how the account entered the system. It over-credits content and ads that create names, and it ignores the webinar that unstuck the deal six months later. Last-touch gives the closer: often a sales email or a branded search. It over-credits whoever was in the room at signature. Both are usable as views. Neither is the truth of a committee sale.

First touch

Gets credit for the intro

Last touch

Gets credit for the form

Reality

A demo plus three emails plus a peer

First and last touch are accounting tricks. The buyer had a messy path.

Multi-touch still assigns guesswork

Linear, time-decay, and position-based models spread credit across captured touches. They look scientific in a BI tool. They still only see what you tagged. Google's Analytics attribution documentation is clear that models are ways of looking at paths, not a photograph of causation. If your CRM never logged the Slack intro from a customer, the model will invent a story without it.

What attribution cannot see

Forwarded PDFs, peer recommendations, a founder talking on a podcast you did not UTM, legal review, and the champion who joined late. B2B deals also have multiple people on multiple devices. Lead-based attribution will split one account into five "customers" or hide four of them. Account-based reporting is less pretty and closer to how you sell.

A practical split

Use last-touch or opportunity-source for sales compensation arguments (keep it boring and stable). Use a simple influenced-account view for marketing: which programs showed up on closed-won accounts. Run incrementality tests on paid when the spend is large enough. Do not rebuild the model every quarter. Model shopping is how teams stop making decisions.

Common questions

Should I buy an attribution platform?

Only after CRM stages are clean and UTMs are consistent. A platform on dirty data will give you confident nonsense. Many teams get further with Salesforce or HubSpot reports plus a monthly won-deal review.

Is "self-reported attribution" worth asking?

Yes, as a second story. "How did you hear about us?" on the form or in the first call catches dark social that pixels miss. It is biased toward the last thing they remember. Use it alongside, not instead of, logged source.

Prefer markdown? B2B marketing attribution, and where it breaks.md