Engagement rate
The share of sessions that lasted, converted, or covered several pages.
In GA4, a session counts as engaged if it lasted longer than ten seconds, triggered a key event, or included at least two page views. Engagement rate is the proportion of sessions meeting any of those conditions.
Key takeaways
- Three conditions joined by or, so a session only has to satisfy one. Ten seconds alone is enough.
- The ten-second threshold is configurable up to sixty. Raise it and the metric falls across the whole property with nothing on the site having changed.
- A key event satisfies the condition on its own, so flagging one raises engagement rate mechanically.
- Bounce rate is its exact inverse, so anything that moves one moves the other by the same amount.
- Check the threshold and the key event list before writing commentary on a jump.
Three conditions, joined by or
A session is engaged if any one of three things happened, not all three. That matters more than it sounds, because it sets the bar at the lowest of them.
The three are: the session lasted more than ten seconds, it triggered a key event, or it included two or more page views. Ten seconds is a very low bar for a human being reading anything, so in practice the time condition does most of the work and the other two are edge cases it catches.
This makes engagement rate a better metric than the old bounce rate and not a measure of quality. A visitor who lands on the wrong page, spends twelve seconds realising it, and leaves is engaged by this definition. Nothing about that visit was any good.
What it is good for is comparison within a property. The absolute number means little; the difference between two landing pages, or the same page before and after a change, means something, because the definition is at least consistent.
The threshold nobody checks
Ten seconds is the default, and it is configurable up to sixty. That setting sits in the property admin, it is invisible in every report it affects, and changing it moves engagement rate across the entire property at once.
The effect is large. Moving from ten seconds to thirty reclassifies every session that lasted between them, and on a content site that is a substantial share of traffic. The chart shows a cliff, the site is unchanged, and the cause is one dropdown.
It also makes cross-property comparison unsafe. Two clients in the same sector with different thresholds are not comparable, and neither is a client property against any published benchmark, since benchmarks rarely state the setting they were computed at.
The habit worth having is to record the threshold alongside the metric the first time it is reported, and to check it before explaining any movement. It is a thirty second check that prevents writing a paragraph about a configuration change.
Why adding a key event moves this metric
The third condition is the one that connects engagement rate to work happening elsewhere in the account. Any session triggering a key event is engaged, so flagging a new key event reclassifies every session that fired it.
If the newly flagged event is common, and outbound clicks or scroll depth certainly are, the movement is immediate and can be large. It appears in the month somebody made the change, which is the same month somebody is writing about what improved.
It runs the other way too. Removing a key event lowers engagement rate, which looks like a regression and is a configuration change.
This is the strongest practical argument for keeping the key event list short and stable. A list that changes every month makes engagement rate, bounce rate and the conversion count all move for reasons unrelated to the site, and month-to-month comparison stops working across all three at once.
What it is actually good for
Engagement rate earns its place at the landing page level, compared against itself over time. Did this page get better after the rewrite. Does this campaign land people who stay. Those are questions the metric can answer.
It does not answer whether a site is good, and a site-wide engagement rate reported monthly becomes a number that moves for compositional reasons: a campaign sending different traffic, a seasonal shift in the mix of pages, a change in the device split. None of those is about engagement.
Segment before drawing conclusions, and segment by the thing that explains the variance. Paid and organic traffic behave differently. Mobile and desktop behave differently. A blended figure hides all of it.
And pair it with something that has commercial meaning. Engagement rate improving while enquiries fall is a real pattern, usually produced by traffic that is more interested and less able to buy, and reporting the first without the second tells a client a story that is not true.
What it means for an agency
If engagement rate jumps between months, check the threshold and the key event list before writing the commentary. Reporting a configuration change as a performance improvement is the kind of error a client remembers.
Where it goes wrong
- Comparing engagement rate across properties, or against a benchmark, without knowing either threshold.
- Reporting a site-wide figure that moves with traffic composition rather than with engagement.
- Letting the key event list change monthly, which breaks engagement rate, bounce rate and conversions together.
- Reporting engagement improving while enquiries fall, which is a real pattern and not a good one.
In short
Engagement rate is the most easily moved metric in GA4 and the least often checked for why it moved. Two settings can shift it across an entire property in an afternoon, so the first question about any change should be whether anything on the site changed at all.
Frequently asked questions
What counts as an engaged session in GA4?
One that satisfies any one of three conditions: it lasted longer than ten seconds, it triggered a key event, or it included two or more page views. They are joined by or rather than and, so the ten-second condition does most of the work in practice, which makes the bar lower than the word engagement suggests.
Why did my engagement rate change when nothing on the site did?
Check two settings. The engagement time threshold is configurable from ten seconds up to sixty, and raising it reclassifies every session in between across the whole property at once. And a key event satisfies the engagement condition on its own, so flagging a common event raises the metric immediately. Both are invisible in the reports they affect.
Is engagement rate better than bounce rate?
It is the same information stated positively, since GA4 bounce rate is defined as its exact inverse. Engagement rate is the better one to put in front of a client because a number going up reads correctly, and because its definition can be explained in a sentence without the history of what the old metric got wrong.
What is a good engagement rate?
The absolute figure means very little, because it depends on the threshold setting, the key event list and the mix of pages and traffic behind it. It is useful compared against itself: one landing page against another, or the same page before and after a change. Published benchmarks are close to worthless here, since they rarely state the threshold they were computed at.
Related terms
Last reviewed
Keplent keeps track of these for you.
It holds what each client agreed to, watches what their campaigns actually did, and tells you in plain sentences what needs you this week.
Join the waitlist