Bounce rate was not removed, it was redefined. GA4 counts a session as engaged if it lasted ten seconds or more, or included two or more pageviews, or triggered a key event. Engagement rate is the share of sessions meeting at least one of those, and bounce rate is simply its inverse. The new definition is better, because the old one described something that was frequently not a failure at all.
What was wrong with the old metric
Universal Analytics counted a bounce as a single-page session with no further interaction recorded.
Consider what that produced. Someone searches a question, lands on your article, reads it carefully for eight minutes, gets a complete answer, and leaves satisfied. Under the old definition that is a bounce, indistinguishable from a visitor who arrived, saw the wrong page, and left in two seconds.
One of those is a success and the other is a failure, and the metric treated them identically. That is why bounce rate produced so many bad decisions: sites added pagination to break articles into multiple pages, inserted interstitials, and fired artificial events, all to improve a number that was measuring the wrong thing.
The new definition
A session counts as engaged if any one of these is true:
| Condition | Detail |
|---|---|
| Duration | Ten seconds or more of foreground engagement time |
| Depth | Two or more pageviews or screenviews |
| Action | At least one key event, meaning a conversion |
These are alternatives rather than requirements, so meeting one is sufficient. Engagement rate is engaged sessions divided by total sessions, and bounce rate is one hundred percent minus that figure. They are arithmetic complements describing the same data from opposite directions.
The eight-minute reader now counts as engaged on the duration condition alone. That is the improvement.
Three details worth knowing
The ten second threshold is configurable
It can be adjusted between ten and sixty seconds per web data stream, under the tag settings for that stream. Most properties never change it, and most owners do not know they can.
Changes apply going forward only, so raising it does not retroactively reclassify anything. For a content site where genuine reading takes longer, a higher threshold produces a more meaningful figure, at the cost of comparability with the period before you changed it.
It measures foreground time
Engagement time accumulates only while the tab is actually in front of the user. A page left open in a background tab does not quietly accrue engagement, which is a meaningful improvement over session duration measures that did.
The two figures are not comparable across platforms
GA4 bounce rate and Universal Analytics bounce rate measure different things, so comparing a current figure against a historical one from the old property is meaningless. Old industry benchmarks do not apply either.
Where it is still imperfect
Worth saying, because a metric treated as authoritative gets misused.
- Ten seconds is arbitrary. A nine-second visit and an eleven-second visit are not meaningfully different experiences, and the boundary decides which is which.
- Engaged is not satisfied. A visitor who spends twelve seconds failing to find something and leaves frustrated counts as engaged. The metric records that something happened, not that it went well.
- Slow pages compress the window. If content takes six seconds to render, the visitor has four seconds to decide before the threshold passes, which links this metric directly to the arguments in how slow before calls drop.
- Some automated traffic qualifies. A crawler that parses a page slowly can satisfy the duration condition, which can inflate the figure on smaller sites.
How to actually use it
Not as a headline number, and not compared against a benchmark from somewhere else.
- Segment by landing page. A reference page with high engagement and no conversions is doing a different job from a service page with the reverse.
- Pair it with conversion rate for the same page and source. The combination distinguishes pages doing useful work from pages that merely detain people.
- Watch it by traffic source. High engagement with no conversions frequently indicates automated traffic rather than an audience.
- Track direction over time on your own site, rather than the level against anyone else's.
The reasonable summary: engagement rate is a better metric than bounce rate was, and it is still a proxy. It answers whether something happened. Whether it was worth happening is a question analytics cannot settle.
Frequently asked questions
Does GA4 still have bounce rate?
Yes, but redefined. GA4 calculates bounce rate as one hundred percent minus engagement rate, so a bounce is simply a session that did not qualify as engaged. This measures something different from Universal Analytics, which counted single-page sessions.
What counts as an engaged session in GA4?
A session meeting at least one of three conditions: lasting ten seconds or more of foreground engagement time, including two or more pageviews or screenviews, or triggering at least one key event. Meeting any single condition is sufficient.
Can the ten second engagement threshold be changed?
Yes, between ten and sixty seconds per web data stream, in the tag settings for that stream. Changes apply going forward only, so raising it does not reclassify historical data, and it costs comparability with earlier periods.
Why was the old bounce rate misleading?
Because it counted a visitor who read an article carefully for several minutes and left satisfied as identical to one who arrived on the wrong page and left in two seconds. Treating a success and a failure the same way produced a great deal of poor decision-making.
Can I compare GA4 bounce rate with my old Universal Analytics figures?
No. The two measure different things, so a direct comparison is meaningless, and older industry benchmarks were calculated under the previous definition and do not apply to the current metric.
How should engagement rate actually be used?
Segmented by landing page and paired with conversion rate for the same page and traffic source, tracked as a direction over time on your own site rather than compared against external benchmarks. High engagement with no conversions often indicates automated traffic.
West Coast Media Solutions Inc. builds websites for businesses across Canada. Platform policies change and enforcement varies; where legal actions are described they are identified by jurisdiction, since much of the published material on these topics is American and does not apply directly in Canada. This is general information rather than legal advice.
Want help getting this right?
We build websites and handle the work that surrounds them, so the details on this page get done properly the first time.
Start a Conversation