B2B marketing attribution fails when three systems tell three stories

B2B marketing attribution is the attempt to answer a simple question: which activity actually contributed to revenue.

In practice the path is long. A lead sees an ad, talks to sales, goes quiet, comes back through a webinar, and closes six months later. Last-click in GA4 credits the last session. The CRM credits the first campaign someone typed in. Finance credits the invoice. None of those are wrong. They are incomplete.

TL;DR

B2B attribution fails when ads, the CRM and billing do not share an identity or a model leadership will actually use. Fix the joins first. Then pick a model (first touch, last touch, linear, or data-driven) and take revenue from the system that invoices, not from the ad pixel. MTA is for journey-level decisions. Mix models are for channel budget. Most growing companies need the first cleaned up before they pay for the second.

The honest summary

A working B2B model usually needs four things:

• A single person and company identity across ads, CRM and billing
• Campaign data that survives form fills and sales-created records
• A stated model that leadership will use in budget conversations
• A revenue number from billing, not from the platform conversion value

Tools only report what entered the CRM. If campaign is blank on half the deals, any attribution product will be confident and wrong.

What actually breaks

Three clocks, three truths

GA4 starts at the session. The CRM starts at the contact. Finance starts at the invoice. Asking them to match without a shared ID is a definitions problem wearing a data problem clothes.

Sales-created records have no campaign

A rep types the opportunity. Source is other or empty. The ad that started the relationship is gone. Native HubSpot and Salesforce attribution reports will still draw a chart.

The ad platform reports its own conversion

Meta and Google will happily show revenue they defined. That number is useful for bidding. It is not booked revenue. Using it in a board pack is how you fund the wrong channel.

MTA vs MMM gets bought in the wrong order

Multi-touch attribution is a join plus a rule on the journey you already have. Marketing mix models need volume and patience. Buying MMM because the CRM is dirty does not clean the CRM.

A model you can defend

Pick one and write it down:

• First touch: useful when you care about how people enter
• Last touch: useful when you care about what closed the week
• Linear: useful when you refuse to pretend one click did the work
• Data-driven: useful only after the identity and campaign fields are trustworthy

Then run the same month through two models once, so leadership can see the difference. The argument you want is which rule, not whose export.

What done looks like

• You can point at an invoice and name the company, the contacts, and the campaigns that touched them
• Campaign is filled on new deals by default, not by a monthly cleanup
• The board pack uses billed revenue, with channel credit from the written model
• Nobody is pasting three CSVs together to make attribution

When this is the wrong project

If you have one channel and ten deals a month, a spreadsheet is enough. If you cannot explain revenue without a weekly reconciliation, you do not have a tool problem. You have a connection problem.