The Checkbox That Changes the Data

Consent mode gets talked about as a compliance switch - flip it on, satisfy the regulation, move on. That framing misses what actually happens on the measurement side. When a visitor declines tracking, the platform doesn't just stop counting them. It tries to estimate what their event would have looked like, and folds that estimate into the same totals as everyone who said yes.

The report still shows one number. It just isn't made of one kind of data anymore.

What "Modeled" Actually Means

The modeling itself is reasonable in principle: use the behavior of consenting visitors who look similar to the ones who declined, and estimate a plausible conversion rate for the group that opted out. It's a statistical fill-in, not a fabrication out of thin air. But it's still an estimate standing in for an observation, and the report doesn't visually distinguish the two.

A conversion count built partly from real events and partly from modeled ones will move for reasons that have nothing to do with the campaign it's supposedly measuring - including a shift in how many visitors decline in the first place.

Where the Modeled Numbers Hide

Most standard reports show a single blended figure with no flag for how much of it is modeled versus observed. That fraction isn't static, either - it moves with the consent rate, which can shift on its own after a cookie banner redesign, a new browser default, or just seasonal traffic mix, none of which have anything to do with actual performance.

A month-over-month jump can be a real change in behavior, or it can be the modeled share of the report growing or shrinking. From the blended number alone, there's no way to tell which one happened.

Reading a Consent-Mode Report on Purpose

Most platforms expose the modeled share somewhere, even if it's not on the default view - a separate report, a diagnostic panel, or an export that breaks observed and modeled conversions apart. Finding that number once is enough to know roughly how much of a given period's total is estimate rather than observation.

The other habit worth building is not comparing periods with meaningfully different consent rates as if they were measuring the same thing - a quarter right after a banner redesign and a quarter six months later can differ in modeled share alone, before performance enters into it at all.

Common Questions

Does consent mode make the data wrong?

Not wrong exactly - the modeled portion is a reasonable statistical estimate, not a fabrication. But it's a different kind of number than a directly observed event, and treating the two identically is where the confusion starts.

How much of a typical report is modeled?

It depends entirely on the site's consent rate, which varies by audience, region, and even how the banner itself is worded - there's no fixed percentage that applies everywhere.

Can the modeled share be turned off?

Not without also losing the visibility into declined-consent visitors entirely - the choice isn't between modeled and exact, it's between modeled and not measured at all.

Where do you actually find the observed-versus-modeled split?

It's usually one layer below the default dashboard - a diagnostics view, a raw export, or a platform-specific report built for exactly this question, rather than the summary screen most people check first.

Not Sure How Much of Your Report Is a Model?

Finding the observed-versus-modeled split usually takes one diagnostic report, not a rebuild.