Attribution: why your channels claim 180% of your revenue

Add up the conversions each ad platform reports and you will find you sold considerably more than you actually did. Nothing is broken. Each platform counts every conversion it plausibly influenced, using its own window and its own rules, and none of them subtract the ones another platform also claimed.

Why the numbers cannot be added

• Different attribution windows — one platform counts a click from 30 days ago, another counts a view from one day ago.
• View-through conversions, which count an impression nobody clicked.
• Modelled conversions, which are estimates filling gaps left by consent and tracking limits. They are reasonable estimates; they are not receipts.
• Last-click analytics, which credits the final touch and ignores everything that created the demand.

The three views worth maintaining

First, platform-reported numbers, used only for optimising within that platform. They are the right signal for the algorithm and the wrong signal for budget allocation.

Second, a single source of truth from your own system — orders or qualified opportunities in your CRM, with a stored first-touch and last-touch source. This is the number that reconciles to revenue, and it is the one finance should see.

Third, incrementality, which is the only view that answers the question you actually care about: what happens if we stop spending here?

Testing incrementality without a research budget

• Geographic holdouts — turn a channel off in comparable regions for a few weeks and compare total revenue, not channel revenue.
• Scheduled pauses on brand search, which frequently show that a large share of that spend was buying traffic organic already had.
• Budget step changes, up or down by a meaningful amount, held long enough to read past the noise.
• New-customer rate by channel, which catches channels that are efficiently reselling to people already coming back.

What to do on Monday

Stop reconciling platform numbers to each other; they will never agree and it is not a data problem. Report from your own system, use platform data only for in-platform decisions, and schedule one incrementality test per quarter on your largest line item. That sequence produces better budget decisions than any attribution model you could buy.