Consent, modelled conversions, and what your bidding actually sees

A quiet change affects a lot of reports, and it rarely comes up in meetings. Some of the conversions you see in an ad account aren't observed events. They're modelled estimates.

That's because some people don't consent to measurement cookies, and others use a browser or device where conventional tracking doesn't work. The platforms fill the gap with an estimate.

Why that isn't a problem in itself

An estimate beats a zero. Without one, the system would decide that a whole segment doesn't convert and move spend away from it. Modelling is what stops that from happening.

Why it still changes your decisions

In three places.

Daily swings are bigger. An estimated number takes longer to settle than an observed one. If you react to daily data, you're reacting to noise.

Segments aren't equal. Where more people decline consent, the estimated share is bigger. So comparing two campaigns isn't always a like-for-like comparison.

Your account and your own system won't match. They don't need to. But if you expect them to, you'll end up treating real data as broken.

What you can do

Get the consent handling right. Do it for measurement too, not only to tick a legal box. If the consent state isn't passed correctly, you either lose more data than you should or collect data you're not allowed to.

Watch your own revenue as well. The ad account shows you the direction. Your books show you the facts. If the two keep drifting apart, you're optimising for the account instead of the company.

Make decisions over a longer window. Looked at weekly or fortnightly, estimation has less effect than it does day by day.

Send more from the server side. Data you send from your own server has fewer things in its way than measurement running in the browser.

The point

The data hasn't gone bad. Part of it is now a probability rather than a count. If you know that, you can make good decisions with it. If you don't, you'll rebuild a campaign over two weeks of variance.