A high bounce rate looks like an alarm in your report, but on its own it settles nothing. In Google Analytics 4 it is the share of sessions that did not meet the engagement criteria. Such a session may come from someone who left after two seconds, or from someone who found your phone number and called straight away. Before you rewrite the headline or rebuild the page, find out which of those situations you are looking at.
This guide does not give you a "good" threshold, because there is none: the result depends on the page type, the traffic source and the definition used by the tool. Instead of a list of tips, you get an order of diagnosis. First you confirm what was actually counted, then you split the number into smaller groups, watch real visits and write a hypothesis for one specific page.
What GA4 counts as a bounce
In GA4, an engaged session lasts longer than 10 seconds, has a key event, or has at least two page or screen views. Bounce rate is the inverse of engagement rate: the share of sessions that met none of these conditions. So a bounce does not mean "left without clicking". It means "analytics recorded none of the three signals in this session".
Google Analytics Help defines an engaged session by three conditions, namely more than 10 seconds, a key event or two views, and bounce rate as the share of sessions that were not engaged. It does not say what a good result is for your industry. Google Analytics Help.
The time threshold can be changed in the data stream tag settings, and whatever you mark as a key event also moves the result. Two sites with the same chart may therefore be measuring different things. Other analytics tools define a bounce in their own way, so compare a page with itself over time, not with a figure from someone else's report. We cover gaps like these in our article on what Google Analytics cannot show you.
Pages where a bounce can mean success
If a page exists to deliver one piece of information, a short visit can mean it worked. A customer checks your opening hours, copies the address, dials the number or reads the answer to a search query and closes the tab. Time spent on page suffers from the same double meaning. The table below helps you separate a bounce with an answer from a bounce with frustration.
Scroll the table sideways to see all columns.
| Page type | Bounce with an answer | Sign of frustration |
|---|---|---|
| Contact | Click on the phone number or email address | Fast scrolling without a single click |
| Article with an answer | Scroll to the paragraph with the solution | Exit after the first screen |
| Offer page | Rarely; the next step is what counts here | Clicks on elements that are not links |
| Form | A submission marked as a key event is not a bounce | Fields started, then abandoned |
Clicks on a phone number are often not measured at all, so a call looks like an exit in the report. We explain how to fix that in our guide to tracking phone number clicks. If you set up that click as a key event for the contact page, some of the "bounces" disappear, because GA4 now counts those sessions as engaged.
Split the number by landing page, source and device
A site-wide average mixes contact pages, blog posts and offers, so you cannot read anything from it. Start with the landing page: a bounce belongs to the first page of the session, not to every pageview. Then, within one page, compare traffic sources and devices. You will often find that a single combination causes the problem while the rest looks ordinary.
- Pick a landing page.Choose a high-traffic page whose job is a next step, not a quick piece of information.
- Split by source.Compare search, campaigns, referrals and direct visits on that same page.
- Split by device.Check mobile and desktop separately; the same layout may only work on one of them.
- Note the worst combination.For example: service page, paid campaign, mobile. This is the group you investigate next.
Watch out for small samples: a few dozen sessions from one campaign can easily produce an extreme result by chance. You can only compare sources reliably if your campaigns are consistently tagged with UTM parameters. When the high share appears only in paid traffic, the problem may lie in the ad's promise, which we discuss in our article on why ad traffic does not convert.
Watch a sample of single-page visits
The number points to a place, not to a reason. For the combination you picked, watch a dozen or so recordings of visits that ended on one page. You are not hunting for one dramatic case but for behavior that repeats. We describe how to select and tag recordings in our guide on how to analyze session recordings.
In Meetlead Analytics you can filter recordings by visited page, number of pages in the visit, device and traffic source, so selecting "one page, mobile, campaign" takes a moment. The customer journey shows the landing and exit page, source, active time and events for each visit. Recordings are available on plans that include them, and whether a visit is recorded depends on your site's privacy settings, such as starting recording only after consent.
- Did the person see the content at all, or leave before the page settled?
- Did they scroll to the answer or the button?
- Did they click something that did not respond?
- Did they scroll back up and wander around as if searching for something?
- Did the exit come right after a click on the number or address?
Check the first screen and scroll reach
Many frustrated bounces are decided on the first screen: the visitor does not see what they came for or cannot tell whether they are in the right place. Check what the group you selected sees on a phone before the first scroll. We list the criteria for that review separately in our guide to your website's first screen.
The second measure is scroll reach. Meetlead records scroll events at 25, 50, 75 and 100 percent of the page, and the heatmap shows clicks and the areas users never reach. If the answer sits in the second half of the page and most visits end before it, you have a concrete hypothesis. We explain how to read this data in our guide to page scroll depth.
Write a hypothesis for one page
A diagnosis ends with one sentence for one page, not a list of ten fixes. A fictional language school saw a high bounce rate on its corporate course page. After splitting the data, it turned out to come mainly from campaign visits on phones. Recordings showed that the price list sat below a long description of the teaching method, and many visits ended before reaching it.
"On [page], visits from [source] on [device] end before [element] because [observation from recordings]. If we move [element] higher, the share of non-engaged sessions in this group will drop over a comparable period." Add the date, the author and what you will accept as counterevidence.
After the change, compare the same group over a similar period, not the whole site. If the share has not moved, the hypothesis was wrong and you go back to the recordings. A lower bounce rate is not a goal in itself either: check whether inquiries, contact clicks or another action you care about went up.
Questions about bounce rate
What is a good bounce rate? There is no single value that fits every page. A contact page with a high share can work perfectly, while an offer page with a low one may not sell. Compare the same page over time and with the same traffic source, and judge the result together with the action the page is meant to trigger.
Will longer copy lower the bounce rate? It may push sessions past the 10-second threshold and improve the number without improving the page. So before you add anything, check in recordings whether people find the answer. If they do and then leave, there may simply be no problem.
