Bounce rate is the percentage of sessions on your site that were not engaged. In GA4, a session counts as engaged if it lasted longer than ten seconds, recorded a key event or included at least two page views; every session that managed none of those is a bounce. That makes bounce rate the mirror image of engagement rate, and the two always add up to 100%.
How bounce rate works
The definition changed with GA4. In Universal Analytics, a bounce was a session with a single interaction, usually one page view, however long the visitor stayed. Someone who read a whole article for five minutes and then closed the tab counted as a bounce. In GA4 that same visit is an engaged session, because it lasted more than ten seconds. On content-heavy sites GA4 bounce rates therefore tend to come out lower than the old ones, and the two cannot be compared.
The ten-second threshold is a default. You can raise it, up to 60 seconds, in the web data stream settings, which makes “engaged” a stricter test. Bounce rate is also missing from GA4’s standard reports until you add it, either through report customisation or in Explorations.
Why it matters
Bounce rate is a quick way to find landing pages that fail the people who arrive on them. If a page about boiler servicing in Croydon receives paid clicks and most of those sessions end within ten seconds, with no second page and no phone tap, then the ad, the search term or the page is wrong. Fixing that page can do more for cost per lead than another round of bid changes.
It works best on pages where you expect a second step: service pages, landing pages for ads, category pages in a shop. On a page whose job is to answer one question quickly, a high bounce rate can mean it worked. That is why bounce rate should be read page by page and next to key events, not as a single sitewide score.
Common mistakes
- Comparing GA4 bounce rate with Universal Analytics history, or with published figures written for the old definition.
- Judging the whole site on one average, which mixes pages built for very different jobs.
- Leaving phone taps untracked. If tapping your number is not a key event, a visitor who found it and rang you within eight seconds counts as a bounce.
- Chasing the number with tricks, such as firing an event on page load or splitting articles across several pages. The figure improves; the site does not.
- Ignoring consent. On a UK site with a cookie banner, bounce rate describes visitors who accepted analytics, and someone who leaves without answering the banner is not counted at all.
How to act on it
Add bounce rate and key events to the Landing page report, filter it to the channel you care about and sort by sessions. For each busy page with a high bounce rate, compare what brought people there, such as the ad text or the search query, with what the first screen of the page shows on a phone. Most fixes come down to keeping that promise: a heading that matches, the price or next step visible without scrolling, and a page that loads quickly.
Make sure every genuine action is tracked first, including phone taps, email clicks and form starts, so the figure reflects real behaviour. Then change one thing at a time and watch the page’s bounce rate and key events over the following weeks. Reworking pages so they answer what visitors came for is the core of on-page SEO, and the same thinking applies to landing pages for paid ads.