Finance wants a number. Last-click will not give you one they trust.

Finance wants a number. Last-click will not give you one they trust.

Every quarter marketing has to show the spend did something. Mix modeling has been the grown-up way to tie revenue to channels. The old version is slow, expensive, and talks about a quarter that is already closed. A model can chew more history, faster, and catch lags a dashboard misses. A TV flight in March that lifts search in May is the kind of thing last-click never sees.

It does not replace the analyst. It lets them answer the CFO questions: revenue per dollar by channel, where you over- or under-spent, what mix to try next.

The file is the model

If spend lives in three sheets with three names for the same channel, the output will sound sure and be wrong. Check three things. Spend by channel, week, and campaign, not a rollup that hides what ran. Sales lined up to the same clock as the media. Outside noise: season, price, competitors. Leave those out and the model hands their effect to your ads.

A language model is useful before the math: scope, variables, holes, hypotheses. "Help me with my mix model" is a shrug. "I have a $40M book across search, linear, CTV, and social, a 30-to-45-day cycle. What belongs in the model, and what is missing?" is a brief.

Show scenarios, not a victory lap

Finance wants ranges, other worlds, and tradeoffs. Baseline last quarter against the outside factors. Move 15% of linear to search and CTV and say what that does to revenue per dollar. Cut 20% and show which channels still work. That is planning, not defending last year's buy.

Do not attribute more than 100% of revenue to media. Name diminishing returns. Run it again next quarter. You do not need a data-science bench. Meridian, Robyn, or a commercial MMM will do if someone can ask a sharp question and throw out nonsense. One quarter, two or three channels, one finding, one recommendation. That beats an eight-month, 60-page PDF nobody opens.