Sessions and rankings do not survive contact with a finance function. Here is the reporting structure that does, and what to do when the numbers disagree.
Marketing reports organic growth. Finance sees revenue that would probably have arrived anyway. Both are looking at real numbers, and the disagreement is structural: they are measuring different things and neither has said so out loud.
Three separations that resolve most arguments
Almost every credibility problem in organic reporting comes from aggregating things that behave differently. Split them and the picture usually settles.
- Branded from non-branded. Branded organic largely reflects demand created elsewhere. Reporting it as an SEO result is the single fastest way to lose a finance team’s trust.
- New from returning. Organic search acquires; it also serves as a navigation layer for people who already know you. Only one of those is growth.
- Assisted from last-click. Report both, consistently, and never switch between them to suit the quarter.
The forecast is the contract
Publish a forecast with a stated range and the assumptions behind it before work begins. It converts every subsequent review from a debate about whether SEO works into a review of a specific prediction, a far more productive conversation, and one finance teams already know how to have.
Include the downside case. A forecast without one reads as a sales document, and finance will price it accordingly.
Proving causation where you can
Where a site has enough regional traffic, geo holdouts are the strongest evidence available: roll a change out in KwaZulu-Natal and the Western Cape, hold Gauteng back four weeks, and compare. Where volume is too low, staged rollouts by template give a weaker but still defensible signal.
A number your CFO cannot reproduce is a number your CFO will eventually stop funding.
One dashboard, two audiences
Marketing needs template-level and cluster-level detail. Finance needs four lines: non-branded organic revenue, cost, contribution and variance against forecast. Build one data model and two views of it. The moment those views come from separate models, they will disagree, and the disagreement will always surface in the meeting you least want it in.