Why does GA4 count different session numbers than Universal Analytics?

GA4 counts sessions differently because it defines them differently: a session only ends after 30 minutes of inactivity, not at midnight and not when new campaign parameters appear, as it did in Universal Analytics. In addition, GA4 estimates the number of sessions statistically, and Universal Analytics often had view filters that GA4 does not have. Google itself states that session-related differences of up to 20 percent between the two systems are not a cause for concern.

Today the question is mostly historical: since July 2024, Universal Analytics data is no longer accessible in the interface or via the API. It still matters when you compare old exports or reports with GA4 figures, for example in a year-on-year view in a slide deck.

What exactly is different about a GA4 session?

• No reset at midnight. In Universal Analytics, a visitor who was on the site at midnight automatically started a new session. GA4 continues the same session.
• No reset on new campaign parameters. In Universal Analytics, a new session began whenever a visitor arrived with different campaign parameters, for example via a second ad or an internal link with UTM tags. GA4 keeps the session running.
• Same timeout. In both systems a session ends after 30 minutes of inactivity by default. In GA4 you can adjust this per web data stream.
• Estimated count. GA4 calculates the number of sessions by estimating the number of unique session IDs (HyperLogLog++). Universal Analytics counted without estimation. In the BigQuery export you can count sessions exactly.

Which factors make the difference bigger?

Google names several factors that influence the gap between the two systems from business to business:

1. Geography. If your visitors are spread over many time zones, more of them cross midnight during a visit. Universal Analytics split these sessions, GA4 does not.
2. UTM tags on your own website. Internal links with UTM parameters restarted the session in Universal Analytics. Sites that used them often show many more sessions in old UA data than in GA4. Google advises against UTM tags on internal links.
3. Filters. Universal Analytics views often excluded traffic, for example internal IP addresses or certain countries. GA4 has no view filters, only data filters such as the one for internal traffic, so the base data differs.
4. Consent. If the consent banner or the Consent Mode setup changed when GA4 was introduced, the measured traffic changes as well. This is not a difference in definition but a difference in data collection.

How should I compare old and new figures?

• Compare within one system. Year-on-year comparisons only make sense if both periods come from GA4. Use the first full year of GA4 data as the new baseline.
• Label old figures. If you show Universal Analytics values in a report, mark them clearly as UA data with a different definition.
• Use more robust metrics. For performance, key events and revenue are more meaningful than sessions. According to Google, purchase counts should match closely between the systems if a unique transaction ID was used.
• Use BigQuery for exact counts. If you need exact session numbers, for example for billing, use the BigQuery export instead of the estimated values in the interface.

Do not treat lower session numbers in GA4 as a loss of traffic. Compare only GA4 with GA4, remove UTM tags from internal links and judge success by key events rather than sessions.

The bounce rate is also based on sessions and has been redefined in GA4. Read more in why the bounce rate disappeared in GA4 and in our article on discrepancies between GA4 and Google Ads.

Sources

• Analytics Help: Comparing metrics between Google Analytics and Universal Analytics
• Analytics Help: About Analytics sessions
• Analytics Help: Google Analytics 4 has replaced Universal Analytics

Not sure whether your GA4 figures are complete? We check tagging, consent and settings and show you which numbers you can rely on. Book a free intro call or read more about our tracking services.