Analytics Reference
GA4 Metrics Glossary
GA4 renamed or redefined most of the metrics people had spent a decade learning, and the official definitions are precise but not always clear. These are the definitions I use when explaining a report to a client.
| Metric | Definition |
|---|---|
| Users | Total distinct users who triggered at least one event |
| Active users | Users with an engaged session or a qualifying engagement signal |
| Sessions | A period of user activity, ended by 30 minutes of inactivity by default |
| Engaged sessions | Sessions lasting over 10 seconds, or with a key event, or 2 or more page views |
| Engagement rate | Engaged sessions divided by total sessions |
| Bounce rate | The inverse of engagement rate, not the old GA4 definition |
| Average engagement time | Time the site was in the foreground, per user or session |
| Event count | Total number of times any event was triggered |
| Key events | Events marked as important, formerly called conversions |
| Dimension | What it tells you |
|---|---|
| First user source or medium | How the user was first acquired |
| Session source or medium | How the specific session started |
| Default channel group | Rule based grouping of source and medium |
| Landing page | First page of the session |
| Campaign | Value passed in utm_campaign or from auto tagging |
| Old term | GA4 equivalent |
|---|---|
| Goal | Key event |
| Conversion | Key event |
| Pageview | page_view event |
| Bounce rate as single page session | Engagement rate, inverted |
| View | Data stream and reporting identity |
Notes
Engaged sessions is the metric to build reporting on, not sessions. It filters out the accidental and bot adjacent traffic that inflates session counts, and its definition is explicit, so you can explain to a client exactly what qualified.
GA4 bounce rate is not the old bounce rate. It is simply one minus the engagement rate, which means a single page visit lasting more than ten seconds is not a bounce. Comparing GA4 bounce rate to a historical Universal Analytics figure is meaningless, and you should say so before someone builds a target on it.
Discrepancies between GA4 and ad platform numbers are expected, not a bug. Ad platforms use their own attribution windows and view through logic, and GA4 uses a different model with different consent handling. Pick one system as the decision source for each question rather than trying to reconcile them.
GA4 renames metrics and changes attribution models more than any analytics platform I have worked with, including the conversions to key events rename. The GA4 dimensions and metrics reference in Analytics Help is the source of record for the current definitions.
Frequently asked questions
What counts as an engaged session in GA4?
A session that lasted longer than 10 seconds, or included a key event, or had two or more page or screen views. The 10 second threshold is configurable up to 60 seconds.
Why do GA4 and Google Ads report different conversion numbers?
Different attribution models, different lookback windows, and Google Ads counts view through conversions that GA4 does not attribute the same way. Both can be correct within their own definitions.
What happened to conversions in GA4?
They were renamed key events. Google Ads still uses the word conversions for what it imports, which is a source of confusion in reporting.
Official sources
Platforms change specifications without notice. Check the primary documentation before a launch that depends on an exact value.
Read next
You might also like
Ranked by how closely each page overlaps with this one, using a similarity model over the whole library.
More Analytics Reference
UTM Parameter Reference
The five UTM parameters, what each one is for, naming conventions that survive scale, and common tagging mistakes.
Conversion Tracking Reference
Conversion tracking methods across GA4, Google Ads and Meta, including client side, server side and enhanced conversions, and when to use each.
Attribution Models Reference
Attribution models compared: last click, first click, linear, time decay, position based and data driven, with guidance on when each is appropriate.