Bounce rate is the share of visitors who land on a page and leave without meaningful interaction, and you reduce it by loading faster, matching what the searcher expected, fixing the mobile experience, and giving people an obvious next step. One thing has to be said before any of that: bounce rate in Google Analytics 4 is not the metric you knew from Universal Analytics. The definition changed, the numbers run lower, and every benchmark written before 2023 now misleads. So this post starts where your diagnosis has to start, with what GA4 actually counts, then maps the fixes to the four causes that produce nearly every bounce, so you treat the disease your data shows rather than working through a generic checklist.
The short version
- GA4 counts a bounce only when a session stays under 10 seconds, fires no key event, and views no second page.
- Engagement rate is GA4's preferred framing; bounce rate is its exact inverse.
- Four causes produce most bounces: slow loading, intent mismatch, broken mobile experience, and no next step.
- Diagnose which cause your data shows before applying fixes; the segments name the culprit.
- +32% bounce probability at 3s load vs 1s
- +123% bounce probability at 10s load vs 1s
- 10s default engagement threshold in GA4
- 60%+ of traffic is mobile, where bounces spike
GA4 redefined the bounce, so recalibrate before you fix anything
Universal Analytics counted every single-page session as a bounce, even a visitor who read your article for five minutes. GA4 flipped the logic. Its primary metric is the engaged session: one that lasts 10 seconds or more, fires a key event, or views a second page. Engagement rate is the share of sessions that qualified, and bounce rate is simply everything that did not, the exact inverse, always summing to 100%. Three consequences follow. First, GA4 bounce rates run far lower than UA ones for the same site, because a 30-second single-page read no longer counts as a bounce, so comparing today's numbers against pre-2023 benchmarks misleads you. Second, the threshold is configurable: an admin can raise the 10-second bar as high as 60 seconds in the data stream settings, which means two sites, or your own site before and after a settings change, can report different bounce rates for identical behavior. Check the setting before trusting a trend. Third, the metric is hidden by default; you add it to reports manually, which is Google's quiet way of saying watch engagement rate instead. We think that framing is right. Optimize for engaged sessions and the bounce number takes care of itself.
Find which of the four causes your data shows
Nearly every bounce traces to one of four causes, and GA4 will tell you which one you have if you segment before treating. Sort pages by bounce rate, then break the worst ones down by device and traffic source, and read the pattern against this table.
| Cause | What the data shows | The fixes that address it |
|---|---|---|
| Slow loading | High bounces across all sources, worst on mobile; poor Core Web Vitals in Search Console | Speed work: images, caching, scripts, INP |
| Intent mismatch | High bounces from organic search only; short engagement time on affected pages | Content and title alignment to the query |
| Broken mobile experience | Mobile bounce rate far above desktop on the same pages | Type size, tap targets, layout stability |
| No next step | Decent engagement time but no second pageview and no key events | Internal links, one clear CTA, related content |
When the cause is speed, every fix is measurable
Google's own research puts bounce probability up 32% at a 3-second load, up 90% at 5 seconds, and up 123% at 10 seconds, each against a 1-second baseline. Nothing else on this page moves the number as reliably as getting under 3 seconds. The work is unglamorous: compress images into modern formats, cache aggressively, serve through a CDN, and defer every script the first paint does not need. Then hold the page to the Core Web Vitals bars, Largest Contentful Paint under 2.5 seconds and Interaction to Next Paint under 200 milliseconds. INP replaced First Input Delay in March 2024, so retire any checklist still naming FID. Speed fixes also compound with everything below, because a visitor who waited too long never sees your improved headline.
When the cause is intent mismatch, the page broke a promise
A bounce from organic search usually means the page did not deliver what the result promised. Three alignment fixes close the gap. Match the content to the query: if the page ranks for a how-to search, the how-to must start on the first screen, not in paragraph nine. Fix titles and metas that overpromise, because a snippet promising what the page cannot deliver manufactures bounces at industrial scale. And answer above the fold, leading with the value instead of a warm-up introduction, since visitors decide in seconds whether to stay. The deeper version of this problem, choosing keywords whose intent your page can genuinely satisfy, is covered in our user intent explainer; ranking for a query you cannot answer buys traffic that leaves immediately.
When the cause is mobile, fix the experience visitors actually get
More than 60% of traffic sees the mobile version first, and mobile is where every small flaw doubles. The mechanical floor: 16px minimum body text, tap targets around 44px, no horizontal scroll, and no layout jumps as images and ads load. Make the content scannable, because walls of text bounce on sight on a phone screen; short paragraphs, honest subheadings, bullets, and white space keep readers moving. And kill intrusive pop-ups. An immediate full-screen overlay is a bounce machine, Google penalizes intrusive interstitials on mobile, and exit-intent or post-engagement timing converts better anyway. Visuals that carry real weight, charts, screenshots, short video, hold attention text alone loses, and they push sessions past GA4's 10-second engagement line.
When the cause is a dead end, give the visit a second act
Some pages load fast, match intent, read well, and still bounce, because they answer the question and offer nothing else. Every page needs a job. Add internal links with descriptive anchors that continue the journey, plus a related-content block at the end so finishers have somewhere to go; every click to a second page is a bounce prevented by definition. Put one clear call to action on every page, chosen deliberately for where the visitor is in their decision. Show trust signals, reviews, client names, security badges, near those decision points so skeptical visitors do not back out at the moment of commitment. And keep the plumbing sound: broken links and 404s end sessions instantly, and stale content with old dates and superseded advice sends readers straight back to the results, so crawl the site quarterly and refresh your best decliners first.
A bounce is not automatically a failure, and bounce rate is not a ranking factor. Blog posts legitimately run high because many readers get their answer and leave satisfied, while a landing page or pricing page bouncing at 80% is losing money by the hour. Compare each page against its own page type on your own site, and track the trend after each fix rather than a single absolute number. The metric is a symptom gauge, not a scoreboard.
Where the metric ends and testing begins
Everything above is the mechanical layer, and most sites get real gains from it. Reading the numbers correctly is its own discipline, from key event setup to segment comparisons, which is what our Google Analytics service handles for clients who inherited a half-configured GA4 property. When bounces on money pages stay stubborn after the mechanical fixes, the problem is usually persuasion rather than plumbing, and opinions stop being useful. That is where structured experimentation takes over. Egochi runs that diagnosis and testing inside its conversion rate optimization service, where bounce analysis is the opening step, not the conclusion.
Questions people ask about bounce rate
What counts as a bounce in GA4?
Any session that was not engaged. An engaged session lasts 10 seconds or longer, fires a key event, or views a second page; everything else is a bounce. The 10-second threshold is adjustable up to 60 seconds in the data stream settings, so two sites can measure the same visit differently.
What is a good bounce rate?
Under the old Universal Analytics definition, 26 to 40% was strong and blog posts naturally ran 65 to 90%. GA4 numbers land far lower because its definition is narrower, so those benchmarks no longer apply. Compare like pages on your own site over time instead of chasing a universal number.
Is bounce rate a Google ranking factor?
No. Google has said your Analytics bounce rate is not used in rankings. The problems behind a high one, slow loading, mismatched intent, and poor mobile experience, hurt rankings on their own, which is why fixing bounces tends to help SEO indirectly.
How do I find bounce rate in GA4?
It is not shown by default. Open a report like Pages and screens, click the pencil to customize, and add Bounce rate as a metric. Engagement rate, its exact inverse, is available the same way and is the framing Google prefers you to watch.
Is a high bounce rate always bad?
No. A visitor who searches a question, reads your answer for eight seconds, and leaves satisfied still counts as a bounce. Judge the number against page intent. A pricing page bouncing at 80% is a problem; a dictionary-style post doing the same may be doing its job.
What is the difference between bounce rate and exit rate?
Bounce rate covers sessions that start and end on the same page without engagement. Exit rate counts how often any page was the last one in a session. A thank-you page can have a high exit rate, a low bounce rate, and be performing perfectly.
