How much of your business ad platforms actually see

It shows up whenever somebody tries to build one number for the business out of the ad accounts. A board pack that reports platform conversions as sales. A blended cost per acquisition calculated by dividing all spend by all platform conversions. A model that assumes if paid says a hundred then the business did roughly a hundred.

It also shows up in the opposite direction, on accounts where organic and referral are strong enough that the ad platforms genuinely account for a minority of what settles, and the paid channel looks far weaker than it is because it is being measured against the whole business.

The two series are answering different questions. A settled payment is all-channel truth: it happened, someone paid, and no attribution rule was involved. A platform conversion is a causal claim about a subset of demand, made under that platform's rules. Comparing them one to one asks a count and a claim to be the same object.

The correlation between them is only stable while conditions are. In months where both platforms ran normally, the two series moved together closely. Add a pause, a promotion, a seasonal peak or a channel that scales, and the relationship comes apart, because the share of demand the platforms can claim has changed even though the business has not.

This is why an unplanned channel pause is so informative. It is the one month where the difference between what the platform claims and what still happens is visible without running an experiment.

Treating the two as interchangeable produces forecasts that work until the first month that is not ordinary, which is the month a forecast is actually needed. Budget plans built on a stable ratio break exactly when a promotion or a pause makes the ratio move.

It also mis-values channels. If platform conversions are read as total sales, every non-paid channel is invisible and paid looks like the whole business. If total sales are read as the paid channel's responsibility, paid looks like it is failing. Both errors get acted on, usually with budget.

Keep both series and never merge them. The settled series answers how much business happened. The platform series answers which levers inside a platform moved. Reporting either one alone is reporting half the picture, and reporting one as if it were the other is reporting a wrong picture.

Track the relationship between them as a series rather than a constant, and expect it to move. When it does move, the useful question is which month was unusual and why, because that is where the causal information is.

And treat any month with a pause, a promotion or a launch as a separate regime. Averaging it in with ordinary months is what turned a correlation of 0.87 into 0.63 on the account above, and the average of two regimes describes neither.

It is not a benchmark for how much of a business paid media should cover. That number is a property of the business model, the brand's existing demand and the maturity of its other channels, and it varies far more between advertisers than the figure above suggests.

It is not an incrementality measurement. Coverage tells you what share of settled outcomes the platforms claim. It does not tell you what share they caused, and those are different numbers.

It is not an argument for reporting only from the payment system. A settled total cannot tell you which keyword, audience or creative to change, which is most of the work.