Explaining causal measurement to your CFO

Marketing and finance often talk past each other. Marketing brings channel reports, engagement and attributed revenue. Finance hears numbers it can't trace, from systems it can't audit, about outcomes it can't verify. Both leave the meeting frustrated.

Causal measurement is the bridge, if it's explained in finance's terms.

Question 1: "What did we get for the money?"

Finance means: compared with not spending it. Answer with incremental results, not attributed ones. "Marketing drove an estimated £4.2m of revenue this year that wouldn't have happened otherwise, most likely between £3.6m and £4.9m" answers the question asked. "Platforms report £9m of attributed revenue" doesn't.

Question 2: "How sure are you?"

Give the range, and say what it rests on. "This comes from a mix model across three years of weekly data, checked against a geo test on our largest channel." A CFO used to forecasting will be more comfortable with an honest range than a precise number they suspect is too precise.

Question 3: "Where should the next pound go?"

Answer with marginal return, not average return. "Paid search has the best average return, but we're near saturation there. The next pound probably goes further in video." Show the expected gain from reallocating at the same total budget, with its range. Finance likes proposals that don't ask for more money.

Question 4: "How would we know if you're wrong?"

This is the question that builds trust. Have a ready answer:

• We test. A geo holdout on the channel we're least sure about, designed before launch, with results read out once.
• We track the forecast. Actual weeks are compared with the forecast range as they arrive.
• We show disagreement. Where two models or a model and a test disagree, we say so.

What to leave out

Platform-reported ROAS, engagement metrics and attribution percentages, unless asked. They answer different questions, and mixing them in invites the suspicion you're trying to avoid.

The shift that matters

Present measurement as a forecast discipline, not a reporting function: estimates with ranges, checked against reality, improved over time. That's language finance already speaks.