Bounce rate

Bounce rate is the share of visits that ended without the visitor doing anything measurable. It is one of the oldest numbers in web analytics, one of the most quoted, and one of the least useful without context. It also does not mean what it used to, because the definition changed when Google replaced its analytics tool, and a great deal of advice still describes the old one.

In short

• It is the share of visits where nothing measurable happened.
• GA4 changed the definition, so old figures and new ones are not comparable.
• A high rate on a page that answers one question is often fine.
• Google says it does not use analytics data as a ranking signal.

The old definition and the new one

The difference is worth understanding once, because almost every argument about this number comes from mixing the two.

The old tool counted a visit as a bounce when it produced only one recorded interaction, which in practice meant a single page view and nothing else. Somebody who arrived, read for ten minutes and left satisfied was a bounce, identically to somebody who arrived and left in a second. That made the number nearly useless for any page designed to answer a question.

GA4 works the other way round. It asks whether a visit was engaged, and a visit counts as engaged if it ran past the engagement timer, produced a key event, or included a second page view. A key event is simply an action you have marked as one of the things the site exists to produce. Bounce rate is simply the share that met none of the three.

The timer is set to ten seconds by default and can be raised as far as sixty, so it is a setting rather than a fixed fact about your site. Since most people who stay to read pass ten seconds, the figure comes out far lower than the old one for the same site.

Why the two numbers cannot be compared

This causes real confusion in reports, so it is worth stating plainly.

A site that showed sixty per cent under the old tool might show twenty five under GA4 with no change whatsoever in how anybody behaved. Nothing improved. The question being asked changed. Any chart that crosses the point where the tools switched is drawing a line between two different measurements, and any benchmark published before then is describing the old one.

The same applies to industry averages, which circulate for years after they stop describing anything. If a figure does not say which definition it used, it cannot be compared with yours.

What a high figure actually means

Almost nothing on its own, which is the honest answer, and it becomes useful when paired with the reason somebody arrived.

• On a page answering one question, a high rate is normal and frequently good. The visitor got the answer and left, which is what the page was for.
• On a landing page behind advertising, a high rate is expensive, because every one of those visits was paid for.
• On a checkout step, it is a problem worth investigating today.
• On a contact page, it usually means something is broken or missing.

The number is a prompt to look, not a verdict. Any interpretation of it that does not mention which page it came from is not saying anything.

It is not the same as exit rate

The two get used interchangeably in meetings and they measure different things.

Bounce rate looks at visits where the page was the only page. Exit rate looks at visits that ended on a page, whether or not anything came before it. So a checkout confirmation page can have a very high exit rate and a very low bounce rate, and both are correct. People arrive at it from somewhere else, then leave, which is exactly what should happen.

The practical difference is which question you are asking. Bounce rate asks whether the page failed to start anything. Exit rate asks whether the page was where people stopped, which on a multi-step process is the more useful question.

One caution before you go looking. GA4 does not report exit rate as a standard metric the way the older tool did, so the way to answer that question now is to build a funnel exploration for the steps you care about.

The changes that move it for the wrong reasons

Bounce rate is unusually easy to move without improving anything, which is why it makes such a poor target.

Marking one more event as a key event will do it, because any session producing a key event counts as engaged. So will raising the engagement timer, which is a setting rather than a fact. The figure falls the following week and nothing about the page has changed. Equally, a page slow enough that people leave before the timer elapses shows a rising rate that is really a speed problem wearing a different costume.

This is the general trouble with a metric that moves for two unrelated reasons. A number that falls when the page genuinely improves, and falls again when somebody changes a setting, cannot tell you which of the two happened.

What to watch instead

The replacement is not another rate. It is a question.

Ask what each page exists to produce, then measure that. A service page exists to produce inquiries. An article may exist to send people to a service page. A contact page exists to produce messages. Once you have decided, the share of visitors doing the useful thing is your conversion rate, and unlike bounce rate it cannot be improved by installing a widget.

Two other things are worth more attention than this number. Scroll depth tells you whether people reached the part of the page that does the persuading. Where they came from tells you whether the page is being judged by the right audience at all, since a page can look like it is failing when it is simply receiving visitors it was never written for.

Bounce rate keeps one job after that. It is a reasonable early warning that something on a page has broken, particularly when it moves sharply on one page while the rest of the site holds steady. Read as an alarm it earns its place. Read as a score it has misled people for twenty years, and deciding what your site should be measuring instead is the whole subject of tracking.