Christian Goodrich published an audit in Search Engine Journal. He took eight UK online stores and compared GA4 with their actual orders.

Last year at one store, between December 25 and January 4, GA4 showed revenue down 68% year over year. The store’s own records showed a 13% drop. At another, GA4 showed revenue down 69%, while sales actually grew 18.5%.

And not a single warning.

What happened?

The cookie banner on these sites ran on a subscription with a monthly cap on views, like a mobile data plan. Black Friday traffic was huge, so the cap ran out in early December. The banner stopped showing until the next billing date, for almost two weeks at one store.

There was nothing to click “Accept” on, so GA4 never saw those shoppers.

The first store received 7,686 online orders that quarter, more than a year earlier. GA4 saw about 41% of them, down from 81%. At the second store, GA4 saw 36% of orders, down from 94%.

Sales were growing. GA4 was showing a decline.

What surprises me most

The bubble I live in.

To me, it’s obvious: if you look at your numbers in the GA4 interface, people declining cookies, and even a broken banner, shouldn’t affect your metrics that much. I’ll explain why below.

Yet big publications, well-known marketers, and brands apparently don’t know this. Which is strange, to put it mildly.

Basic vs Advanced Consent Mode

Google has two Consent Mode options: Basic and Advanced.

In Advanced mode, tags still fire when a user declines cookies. They are sent without cookies or personal data, but they are still sent. That’s the key part.

GA4 then uses these anonymous signals to model the behavior of people who declined consent and includes them in reports. It’s an estimate, but in practice it can be very close to reality.

There are limits. You can’t properly access these users in raw data. Modeling doesn’t start immediately: GA4 needs enough events from users with and without consent.

But if you work primarily in the GA4 interface rather than raw data, Advanced Consent Mode shouldn’t wreck your data across sources, devices, and other dimensions.

At least not on the scale of 50–55%.

All of this is in Google’s documentation. There are plenty of videos and posts about it.

That’s why stories like this leave me stumped. Beginners or people far from marketing and analytics might not know it. But people who write about this professionally or consult on it?

I don’t know.

The good news: as a fractional head of analytics, I’m clearly not running out of work. Right before Black Friday, I’ll take that.

Source: Christian Goodrich’s audit in Search Engine Journal.

Want to get more out of GA4 and BigQuery with AI? Grab the free AI Playbook for the Product Analyst: prompts, workflows and a 90-day rollout plan.