There is no single good bounce rate. Contentsquare's benchmark data puts the cross-industry average at 43% on desktop and 51% on mobile, and 47% for ecommerce.
A content page or single-step landing page can run much higher and still do its job, so a "good" bounce rate is one that falls over time for the same page, traffic source and device, while revenue holds or rises.
For conversion work, bounce rate is a symptom to investigate, not a goal to optimize.
The tests in our database show pages that cut bounces without selling more, and pages that sold more while engagement numbers barely moved.
Bounce rate vs exit rate
Bounce rate and exit rate both describe people leaving, but they count different things.
Google's original Universal Analytics help page defines them like this:
- Bounce rate for a page: of all sessions that started on that page, the share where it was the only page viewed.
- Exit rate for a page: of all pageviews of that page, the share that were the last in the session.
So bounce rate is only about entrances. Exit rate counts every visit to the page, wherever the person came from.
Every session has to exit somewhere, so a high exit rate on its own tells you very little.
A worked example
Google uses five sessions to show the difference:
| Day | Path |
|---|---|
| Monday | Page B, Page A, Page C, exit |
| Tuesday | Page B, exit |
| Wednesday | Page A, Page C, Page B, exit |
| Thursday | Page C, exit |
| Friday | Page B, Page C, Page A, exit |
Take Page B. It was viewed 4 times (Monday, Tuesday, Wednesday, Friday) and was the last page twice (Tuesday, Wednesday), so its exit rate is 2 / 4 = 50%.
Three sessions started on Page B (Monday, Tuesday, Friday) and one of them went nowhere else (Tuesday), so its bounce rate is 1 / 3 = 33%.
The full results:
| Page | Exit rate | Bounce rate |
|---|---|---|
| Page A | 33% | 0% |
| Page B | 50% | 33% |
| Page C | 50% | 100% |
Page C has a 100% bounce rate from a single entrance.
That is the first lesson about bounce rate: on low-traffic pages it swings wildly, and one or two sessions can make a page look broken.
When each one is useful
- Use bounce rate on pages where people land from ads, email or search: landing pages, the homepage, collection pages. It tells you whether the first impression made anyone take a second step.
- Use exit rate on pages inside a flow, like cart, shipping and payment steps. A rising exit rate on a checkout step is a leak in your conversion funnel; a high exit rate on an order confirmation page is expected.
- Use neither as the deciding metric of an A/B test. Both are proxies. The test should be judged on conversions or revenue per visitor.
Bounce rate: what it measures and when it matters
In Universal Analytics, Google defined bounce rate as "single-page sessions divided by all sessions".
The same page says the obvious part out loud: if your site depends on people viewing more than one page, a high bounce rate is bad, but for a blog or other content where single-page visits are expected, a high bounce rate is "perfectly normal."
That old definition had a known flaw. Someone who read a 2,000-word article for six minutes and left counted as a bounce, exactly like someone who closed the tab after one second.
GA4 fixed this by redefining the metric, covered in the next section, so GA4 bounce rates are usually lower than the old Universal Analytics numbers for the same site.
Benchmarks, with caveats
Contentsquare publishes the most widely used cross-industry figures. Its bounce rate write-up gives these averages:
| Segment | Average bounce rate |
|---|---|
| All industries, desktop | 43% |
| All industries, tablet | 45% |
| All industries, mobile | 51% |
| Ecommerce overall | 47% |
| Ecommerce, grocery (highest of nine sectors) | 59.03% |
| Ecommerce, sports (lowest of nine sectors) | 49.80% |
Treat these as a rough frame, not a target.
They come from Contentsquare's own tracking, not GA4, and bounce rate depends heavily on how a tool defines a bounce, which traffic you send, and whether you count single-page apps correctly.
A paid social campaign aimed at cold audiences will bounce more than branded search, on the same page.
A better benchmark is your own history. Segment bounce rate by landing page, channel and device, then compare each segment against itself month over month.
A product page whose mobile bounce rate jumps from 45% to 60% after a theme update is worth a look.
A product page that has always sat at 60% is not, by itself, a problem.
What our A/B tests show about bounce rate
Few teams make bounce rate the primary metric of a test, and the ones in our database that did mostly had small samples or no sales data. That pattern is itself a lesson.
- MedaliaArt, an online painting gallery, moved its Holiday Sale message from small sidebar text to big red type on the homepage. Bounce rate fell from 76% to 60% and clicks rose from 24% to 40%. The test reached 95% confidence, but on only 268 visitors (145 vs 123), and nobody measured whether more paintings sold.
- Abila, a nonprofit software company, personalized its homepage banner by segment: customers saw support and training, partners saw the marketplace, prospects saw current campaigns. Time on site rose 171%, pages viewed rose 107%, and bounce rate fell 85% for targeted visitors. This was a before-and-after comparison, not a controlled test, and it reported engagement, not leads.
- OnSwitch, a freemium lighting app, locked some popular free scenes for the first three days and unlocked them with a push notification. Users came back slightly more but bought less, and a repeat with different locked content gave similar results. Better engagement, worse business.
The OnSwitch result is the one to remember.
Any change that makes people click around more can lower bounce rate: hiding information behind tabs, splitting a page in two, adding a pop-up that counts as an interaction.
None of that guarantees a sale. If you want bounce rate in a test, use it as a guardrail metric or a diagnostic, and keep the decision on revenue.
How to lower a bounce rate that really is a problem
When a landing page bounces far more than its own history or its sister pages, the causes are usually plain:
- Message mismatch. The ad promised one thing and the page opens with another. Our ad message match tests are the best place to start.
- Slow or broken first load, especially on mobile. Check bounce rate by device and browser before redesigning anything.
- The offer is hidden. MedaliaArt's sale existed all along; it was just in the sidebar.
- The wrong visitors. No page fix helps traffic that was never going to buy. Fix targeting first.
Pair the numbers with heatmaps and scroll maps to see where people stop, then test the fix and judge it on conversions.
Engagement rate in GA4
GA4 builds bounce rate on top of a different idea: the engaged session.
Google's engagement rate documentation defines an engaged session as one that meets any of these three conditions:
- Lasts longer than 10 seconds.
- Has a key event (what GA4 used to call a conversion).
- Has 2 or more page or screen views.
From there the formulas are simple:
- Engagement rate = engaged sessions / total sessions
- Bounce rate = 1 - engagement rate (the share of sessions that were not engaged)
Worked example: a page has 2,000 sessions in a month.
1,240 of them lasted more than 10 seconds, fired a key event or reached a second page.
Engagement rate is 1,240 / 2,000 = 62%, and bounce rate is 38%.
Under the old Universal Analytics definition, a reader who spent three minutes on that page and left would have been a bounce. In GA4 they count as engaged.
Things that trip people up
- The 10-second timer can be changed. GA4 lets you set the number of seconds before a session counts as engaged in the data stream's tag settings (Google's session settings help). Raising it lowers your engagement rate overnight without any real change in behavior, so note the date if you change it.
- Bounce rate isn't always in the default reports. You may need to add it to a report or an exploration. Our GA4 for CRO guide covers the setup.
- Key events inflate engagement. If you mark a cheap event, like a scroll or a video start, as a key event, nearly every session becomes engaged. Keep key events for actions that matter, and track smaller steps as micro-conversions.
- Do not compare GA4 and Universal Analytics bounce rates. They measure different things. A drop after migration is a definition change, not an improvement.
Engagement metrics in A/B tests
Several tests in our database used engagement as the primary metric.
They are useful signals, but read the caveats:
- Medienreich Training, a German software training provider, replaced three broad training formats on its homepage with eight best-selling course topics. Engagement rose 40.87% at 99.9% significance over about 20 days. Enquiries were not the primary metric.
- UK Tool Centre removed the filter menu from a small woodcare category page and saw engagement rise 27%, with no sales figures, sample size or significance reported.
- Encyclopaedia Britannica tested other colors for in-article links and found non-blue colors cut link engagement by about 10% on average. For a content site, clicks on links are close to the real goal, so this one is a fair use of engagement.
Compare those with a test judged on money.
Clean Commit's homepage redesign for a Shopify store showed +8.0% conversion rate and +8.6% revenue per visitor, but on 5,092 visitors with a 76.8% probability to beat control, and desktop revenue per visitor fell 12.9%.
The agency called it a positive signal, not a win.
Engagement-style improvements on a homepage often look like this once revenue is in the picture.
The practical rule: watch bounce rate and engagement rate to find pages worth investigating, then run the test on conversions or revenue per visitor with enough traffic to trust the answer.
If your traffic is thin, our guide to testing on low traffic explains when a proxy metric is a reasonable compromise.
To see what has actually moved revenue on pages like yours, browse the full test database.