GA4 shows more users than Varify
The symptom
GA4 shows more users, sessions or events (conversions) than Varify.
This is not an error. With Consent Mode v2, GA4 and Varify count different groups of visitors, so the numbers are not meant to match exactly.
Why the numbers differ
With Consent Mode v2 active, the sources count different groups:
| Source | What is counted | Effect | Example |
|---|---|---|---|
| GA4 interface (Blended) | Users with consent plus modeled users without consent | Highest numbers, because estimated values are included. Not directly comparable with Varify. | approx. 8,500 users (6,000 with consent + approx. 2,500 modeled) |
| GA4 audiences (imported into Varify) | Users who belong to an audience (based on user IDs) | Contains only users with consent and no modeled data. | approx. 6,000 users |
| GA4 raw data (BigQuery) | Events and users from the GA4 export | Contains only users with consent and no modeled data. | approx. 6,000 users |
| Varify Realtime Tracking | Users who are tracked after varify.setTracking(true) | Contains only users with consent and no modeled data. | approx. 6,000 users |
Example values are made up to illustrate the order of magnitude (assumption: 10,000 visitors, 60% consent rate).
Why GA4 is higher
For users who did not give consent, GA4 estimates users, sessions and events based on the behavior of comparable users who did consent. These modeled values only appear as aggregated numbers in the GA4 reports. They are not individual, real users, and you will not find them in audiences or in the raw data.
Why Varify only counts real users
An A/B test can only test users who are actually on the page, are assigned to a variation and can be measured. Modeled users are a statistical estimate. You cannot show them a variation, and their behavior cannot be assigned to a variation.
That is why Varify only uses data from users with consent. This applies to Varify Realtime Tracking and to imported GA4 audiences.
Conclusion
It is completely normal that the GA4 interface shows higher numbers than Varify. In A/B testing, only real users with consent are tested, never modeled ones. Varify, audiences and BigQuery are on the same level and can be compared directly with each other.