Your Modeled Conversions Are a Guess. Check Whether Google Is Even Modeling.

Modeled conversions only exist in Advanced mode, where tags send cookieless pings, and only above strict volume thresholds. If your site runs Basic mode or sits below 700 ad clicks in 7 days, Google is not modeling anything, and the numbers you see are not modeled, they are missing.

Basic mode has nothing to model from

Consent Mode v2 comes in two modes, and the difference between them decides whether modeling is even possible. Basic mode blocks all Google tags until a visitor grants consent. If they reject, no tags fire, no data is collected, and Google has zero signals to work with. Modeling under Basic mode relies on broad industry benchmarks rather than your own site behavior, which makes it far less accurate, and for many purposes it is not modeling at all, it is a generic estimate wearing your brand name.

Advanced mode loads Google tags on every page regardless of the consent choice. When consent is denied, those tags send what Google calls cookieless pings, limited signals that carry coarse information like a random number per page load and a boolean consent state. Those pings are what feeds modeling. Without them, there is nothing to model from, and the modeled conversions in the report are a fiction. The mode choice is therefore not a technical detail. It is the decision that determines whether the numbers in the report mean anything.

The thresholds, stated plainly

Modeling does not activate automatically, and the thresholds are strict. For Google Ads conversion modeling, the requirement is 700 ad clicks over a 7-day rolling window, per country and domain grouping. For GA4 behavioral modeling, the requirement is at least 1,000 events per day with analytics storage denied for at least 7 days, and at least 1,000 daily users sending events with analytics storage granted for 7 of the previous 28 days. Meeting the threshold once is not enough. It is a continuous requirement, and if the volume drops below it, the modeling stops.

Most small and mid-size stores fail these thresholds, which means most small and mid-size stores have no modeling at all, whether they know it or not. A site receiving fewer than 1,000 daily sessions from users who deny consent will never see GA4 behavioral modeling activate. A site with fewer than 700 ad clicks in a week will never see Google Ads conversion modeling. The modeled conversions those sites report are not modeled. They are the observed conversions only, and the unobserved ones are simply absent, which is a different and worse problem than an inaccurate model.

What the model assumes about your visitors

The model works by comparing two populations: visitors who consented, whose full journey is tracked, and visitors who denied consent, who sent only cookieless pings. It identifies behavioral patterns, pages viewed, time on site, device type, geography, referral source, that correlate with conversions among consented users, and it applies those patterns to the unconsented group to estimate a conversion probability for each session. Google has stated that this recovers, on average, more than 70 percent of the ad-click-to-conversion journeys that would otherwise be lost.

The assumption underneath is that consented users are reasonably representative of unconsented users, and that assumption is the weak point. People who consent and people who refuse are different populations, with different behaviors, and the model has no way to know how different, because the unconsented group is the one it cannot observe. Google does not publish confidence intervals or error margins for modeled conversions, so there is no built-in way to gauge how reliable a specific number is. The modeled total is an estimate, and it should be reported as one, with the caveat attached, not as a measured fact.

The check before you move budget

Before moving budget on modeled numbers, check three things. First, the mode: confirm the implementation is Advanced, not Basic, because Basic mode has nothing to model from. Second, the thresholds: confirm the property meets the 700-click or 1,000-event minimums, because below them there is no modeling. Third, the validation: run a holdout test, or compare the modeled total against a media mix model, before treating it as real. Modeled conversions are estimates of observed-versus-unobserved journeys, and they should be validated with independent measurement before they move money. A budget moved on an unvalidated model is a budget moved on a guess, and the guess is wearing a number that looks like a measurement.

Frequently asked questions

What is the difference between Basic and Advanced Consent Mode?

Basic blocks all Google tags until consent is granted, so there is nothing to model from when users refuse. Advanced loads tags on every page and sends cookieless pings when consent is denied, which is what feeds modeling.

What volume does Google require before modeling starts?

For Google Ads, 700 ad clicks over 7 days per country and domain grouping. For GA4 behavioral modeling, 1,000 events per day with analytics storage denied for 7 days, plus 1,000 daily granted users for 7 of the previous 28 days.

How accurate are modeled conversions?

Google states modeling recovers on average more than 70 percent of lost ad-click-to-conversion journeys, but it publishes no confidence intervals or error margins. Treat the number as an estimate, not a measurement, and validate it before moving budget.

Does modeling restore the lost user journeys?

No. It estimates aggregate outcomes, not individual journeys. You will not recover a named user path, only a modeled count, and the model assumes consenting users represent the unconsented group.