Bounce rate is the percentage of sessions in which someone lands on the site and leaves without doing anything else.
The thing to know before opening the report: in Google Analytics 4 this metric no longer means what it meant in Universal Analytics. It is now the exact complement of engagement rate. If engagement rate is 62%, bounce rate is 38%, and there's no other calculation behind it. That makes any comparison with data collected before the move to GA4 meaningless, as it does with most of the benchmarks still doing the rounds.
How bounce rate is calculated
The formula is sessions without engagement divided by total sessions, multiplied by a hundred. Out of 1,000 sessions, if 380 met none of the engagement criteria, bounce rate is 38%.
In GA4 the metric doesn't appear in the default reports. You have to add it by hand: open a detail report, for example Pages and screens, click Customize report at the top right, then Metrics, Add metric, and search for Bounce rate. You need the Editor or Administrator role, and the option isn't available in overview reports. If you can't see the button, it isn't a configuration problem: you don't have the permissions.
What counts as an engaged session
This is the point almost every guide gets wrong. The Google Analytics documentation defines an engaged session as one that meets at least one of these three criteria:
- it lasts longer than 10 seconds
- it has a key event
- it has two or more page or screen views
Nothing else counts. Scrolls, outbound link clicks and PDF downloads are events GA4 collects through enhanced measurement, but on their own they don't turn a session into an engaged one unless that event has been marked as a key event. The line you often read, that a bounce is anyone who makes no interaction at all, scrolling included, describes a behaviour GA4 doesn't have.
The same support page includes an explicit exception: the first_visit, first_open and session_start events are left out of the engaged sessions count even when they are marked as key events.
The practical consequence: on a property with no key events configured, bounce rate measures one thing only, namely how many people stayed less than ten seconds on a single page. It doesn't measure content quality.
You decide the ten-second threshold
The ten seconds are quoted everywhere as if they were a constant. They are a default, and you change them in the interface: in the Google tag settings guide the option is called Adjust timer for engaged sessions.
| Setting | Default value | Where to change it | Effect on bounce rate |
|---|---|---|---|
| Timer for engaged sessions | 10 seconds | Admin > Data streams > Configure tag settings | Raising it pushes bounce rate up, lowering it brings it down |
| Session timeout | 30 minutes | Same path | A shorter timeout splits visits into more sessions and usually raises the figure |
| Key events | None active by default | Admin > Key events | Every extra key event moves sessions from bounces to engaged |
| Enhanced measurement: scrolls, outbound clicks, downloads | On | Configure tag settings | None, until those events become key events |
Two GA4 properties on the same site, with different timers and different sets of key events, return different bounce rates for exactly the same users. Before comparing your figure with anyone else's, it's worth knowing which threshold they were using.
Why the benchmarks you find online no longer hold
The industry tables on the first page of results nearly all come from the same source: CXL's collection of benchmarks, built on Google Analytics benchmarking data when the metric was still the Universal Analytics one. Values run from 44.50% for real estate to 65.52% for food and drink, on a US sample.
They are readable figures, but they measure something else. In Universal Analytics a bounce was anyone who sent a single request to the Analytics server, and those sessions had zero duration by definition. In GA4 a bounce is anyone who meets none of the three criteria. A page read for two minutes without a click was a bounce in Universal Analytics and isn't in GA4.
The same applies to the 30-second convention that the Italian Wikipedia reports, citing Paul W. Farris's Marketing Metrics: it's a historical reference from older web statistics systems, not a GA4 criterion.
The only comparison that holds is with yourself: same site, same property, same settings, different periods. On a B2B manufacturing project we worked on, for instance, the metric that told the story of the result wasn't this one but average position and share of voice: the figures are in the case study on SEO for manufacturing.
Bounce rate, exit rate and pogo-sticking
| Metric | What it measures | Where to find it | When it's really useful |
|---|---|---|---|
| Bounce rate | Percentage of sessions without engagement | GA4, added to the report by hand | Checking tracking and the match between ad and page |
| Exit rate | Percentage of views of a page that were the last in the session | GA4, page reports | Finding where a multi-step journey breaks off |
| Pogo-sticking | A quick return to the SERP followed by a click on another result | No tool exposes it as a metric | Thinking about the match between search intent and content |
The difference between the first two is clearest with an example. Someone who visits page A, then B, then C and leaves from C produces an exit on C and no bounce. Someone who opens only A and closes the browser after three seconds produces both an exit and a bounce on A.
Bounce rate is not a ranking factor
Google's John Mueller said so in a webmaster hangout, reported by Search Engine Roundtable: there is a misconception that Google looks at things like the Analytics bounce rate when ranking websites, and that is definitely not the case.
That doesn't make it a useless metric. It means there's no direct lever: lowering the number doesn't produce rankings. The causes that often sit behind a high figure do matter, but they are measured elsewhere, through page speed, the match between query and content, mobile performance and, more generally, the user experience of the landing page.
In the UK there's one more filter: consent
Bounce rate is calculated on tracked sessions, not on all visits. Anyone who rejects cookies, or closes the banner without choosing, isn't counted if the tag stays blocked until consent. The denominator is already a subset, and not a random one: people who accept at the first click behave differently from people who refuse.
Since 5 February 2026, the Data (Use and Access) Act 2025 has added a statistical purposes exception to PECR, explained in the ICO guidance on storage and access technologies. Analytics cookies can be set without prior consent when their sole purpose is to collect statistics on how the website is used with a view to improving it. The conditions are specific: clear and comprehensive information about the purpose, a simple and free way to object, data aggregated so it can't identify individuals and, if an analytics provider is involved, a provider acting only as your processor and not combining the data with its other sources.
The exception covers statistics and nothing else: as soon as the same data feeds advertising or profiling, prior consent is needed again.
Put simply: two UK sites with the same audience and two different consent setups, one relying on the exception and one still waiting for opt-in, will read two different bounce rates. Before blaming a high figure on content quality, it's worth knowing what share of traffic it's calculated on.
What to look at before touching the page
An aggregate bounce rate tells you nothing. It needs segmenting, in the order that takes least time:
- by landing page, to see whether the problem is widespread or limited to one template
- by device: a clear gap between mobile and desktop almost always points to a rendering or speed problem
- by channel, because paid and social traffic bounce more than organic even on sites that work
- by period, putting the curve next to the date of the last change to the site or the tag
If the figure is below 10%, or exactly zero, the problem isn't the site: it's the tracking, and it's almost always a tag installed twice. If instead it has shot up on a single group of URLs, before rewriting the copy it's worth checking that those pages still exist and aren't returning a 404 error.
The most useful way to use this metric is not as a target to improve but as an alarm. A sudden jump on a single template after a release shows up in bounce rate before it shows in conversions and well before it shows in rankings: that's why it belongs among the monitoring SEO KPIs rather than the outcome ones. If you need a hand working out which numbers really matter on a project, we cover it in our SEO services.
Bounce rate: frequently asked questions
There is no value that suits everyone, and benchmarks published before GA4 can't be compared with current ones. The only sensible comparison is with the history of the same property, with unchanged settings. On a contact page or an article that answers the question straight away, a high figure is normal.
No. Google's John Mueller has stated that the Analytics bounce rate is not used to rank websites. The causes that push it up, such as slow loading or content that misses the intent, can have an indirect effect on rankings.
In Universal Analytics a bounce was anyone who sent a single request to the Analytics server. In GA4 bounce rate is the complement of engagement rate: a bounce is anyone who stays less than ten seconds, triggers no key events and views fewer than two pages. Historical data from the two platforms can't be compared.
Hardly ever. A zero or very low figure is usually a tracking error: a tag loaded twice sends several requests per session and wipes out bounces. Check the installation before treating it as a result.
It doesn't appear in the default reports. Add it to a detail report through Customize report at the top right, then Metrics, Add metric, Bounce rate. You need the Editor or Administrator role.
