Engagement
What is bounce rate, and should I worry about it?
The definition changed and most of the internet did not notice. Knowing which version you are looking at decides whether the number means anything at all.
What is bounce rate and should I worry about it?
In GA4, bounce rate is the percentage of sessions that were not engaged, which makes it the exact inverse of engagement rate.
A session counts as engaged if it lasted longer than ten seconds, recorded a key event, or included at least two page or screen views, so a bounce is a visit that did none of those three things.
On its own it is a weak signal and rarely worth chasing, and it only deserves attention when one page bounces far harder than comparable pages on the same site.
Bounce rate is engagement rate, turned upside down
GA4 calculates engagement rate first. Bounce rate is then simply everything left over: one hundred per cent minus the engagement rate, expressed per session.
That is a genuinely different measurement from the one most people carry in their heads, and the difference matters. Under the old Universal Analytics, a bounce was a session with exactly one interaction. Someone could land on your opening hours page, read it for nine minutes, get the answer they needed and leave satisfied, and the report counted that as a bounce with a session duration of zero seconds. The metric was structurally unable to tell a satisfied reader from a frustrated one.
Read the full breakdown: Bounce rate is engagement rate, turned upside downHide the full breakdown: Bounce rate is engagement rate, turned upside down
GA4 fixed part of that. It measures engagement time in the foreground, so a page that genuinely gets read usually clears the ten second bar by itself and never becomes a bounce. What remains in the bounce bucket is closer to what people always assumed it measured: visits where nothing happened at all.
One practical note before you go looking. Bounce rate is not in GA4's default reports. You add it by customising a report's metrics, or by building a free-form exploration in Explore. Engagement rate is the one you will see first, and it is the same information.
The three ways a session counts as engaged
Any one of these is enough. A session that meets none of them is a bounce, and a session that meets all three is counted once.
- It lasted longer than ten seconds of engagement time, measured while the tab was actually in the foreground.
- It recorded at least one key event, which is the name GA4 now uses for what it used to call a conversion.
- It included at least two page views or screen views.
- The ten second threshold is adjustable. It sits in the data stream settings, under the tag configuration, and can be raised if ten seconds is too low a bar for your content.
That last point is the one people miss. If your bounce rate moved sharply and nobody changed the site, check whether somebody changed that timer.
It is not a ranking signal, and it never quietly was
Bounce rate is calculated inside your own analytics property, from data your own tag collected. Google has said repeatedly that it does not use Google Analytics data in search ranking. Treating bounce rate as a lever you pull to rank better is a category error, and a fair amount of advice on this topic is built on it.
There is a real idea underneath the myth, which is why it survives. A page that answers the question it promised tends to keep people, earn links and get returned to. That is worth pursuing. Pursuing it by trying to move a number in a dashboard is not the same activity, and it leads to genuinely bad decisions such as splitting a clear one page answer across four pages purely to force a second view.
There is also a boring technical cause behind a surprising share of alarming bounce rates: a tag installed twice, once directly and once through a tag manager. Two installations produce duplicate page views, which means every session has two views, which means almost nothing is ever a bounce. If your engagement rate looks implausibly good, check the installation before you congratulate anyone.
Sometimes the bounce happened before the click
A results page now answers a great deal on its own, and that changes who is left to arrive at your site.
This cuts both ways for a bounce rate. Visitors who arrive having already read a summary tend to be further along and more likely to act, which pushes engagement up. Visitors who arrive still looking for a single fact will take it and go, which pushes engagement down and tells you nothing bad about the page.
The practical consequence is that you cannot compare bounce rate across pages with different jobs. A location page whose job is to give an address and a phone number should have a high bounce rate. A service page whose job is to produce an enquiry should not. Comparing the two produces a conclusion that is worse than no conclusion.
Compare a page against itself over time, or against pages with the same job, or do not compare it at all.
How to read a high bounce rate without panicking
If one page really is losing people, this is the order that finds the cause fastest and avoids the usual wasted rebuild.
Split it by landing page first
A site-wide bounce rate is an average of pages doing unrelated jobs and it is close to meaningless. Build the report by landing page so you are looking at one page's behaviour rather than a blended figure.
Split again by channel and device
A page that holds desktop visitors and loses mobile ones has a layout problem. A page that holds organic visitors and loses paid ones usually has a promise problem: the ad said something the page does not deliver.
Ask what the page is actually for
Write down, in one sentence, the single action you want from this page. If that action is a phone call from a mobile, a bounce is a customer and the metric is lying to you. Fix the measurement, not the page.
See the remaining steps: How to read a high bounce rate without panickingHide the remaining steps: How to read a high bounce rate without panicking
Check speed and layout stability
Slow first loads and content that jumps while it renders both cause genuine abandonment. The thresholds published by Google are 2.5 seconds or less for Largest Contentful Paint, 200 milliseconds or less for Interaction to Next Paint, and 0.1 or less for Cumulative Layout Shift, measured at the 75th percentile and split by mobile and desktop.
Verify the tag before rebuilding anything
Confirm the page fires one page view, not two, and that any form, call or chat on it is recorded as a key event. A large share of bounce rate emergencies are measurement faults that cost nothing to fix.
Change one thing, then wait
Change the headline, or the position of the call to action, or the load time. One of them, not all three. Give it enough traffic to mean something before drawing a conclusion, which on a small local site is usually weeks rather than days.
If all six come back clean and the page still bounces, the honest answer is often that the page ranks for a question it does not answer. That is a content problem, not an engagement problem.
Better numbers to spend your attention on
Bounce rate is a symptom metric. These three are closer to the decisions you actually make.
What is a good conversion rate?
The number bounce rate is usually standing in for, and a more honest way to judge whether a page works.
What is a good conversion rate?What is conversion rate optimisation?
How to test a page change properly instead of redesigning on the strength of one bad-looking metric.
What is conversion rate optimisation?What is a landing page?
Why a page built for one action behaves so differently from a general page, in the reports and in real life.
What is a landing page?The web design guide
Speed, structure and layout stability, and how each one shows up in the way visitors behave.
The web design guidePages earn a second action, they do not trap one
The reliable way to improve engagement is to make the next step obvious and worth taking.
Give the answer in the first screen. A page that makes people hunt for the thing they came for loses the ones who are in a hurry, and in a local search almost everyone is in a hurry.
Then offer one next step, not six. A satisfied reader will take an obvious, relevant link. A confused reader takes none of them. Six competing calls to action at the bottom of a page is the most common version of this mistake.
Make contact possible from wherever the reader currently is. A tap-to-call link that works on a phone, a short form, a booking link. If the only way to contact you sits at the bottom of a long page, your engagement problem is really a layout problem.
The short list of times bounce rate is genuinely telling you something
Most of the time the answer is no, do not worry about it. These are the exceptions worth acting on.
- One page bounces far harder than other pages doing the same job on the same site.
- A page's bounce rate jumps sharply on a date you can point at, which usually means a deployment, a tag change or a speed regression.
- Paid traffic bounces much harder than organic traffic on the same landing page, which points at a mismatch between the ad and the page.
- Mobile bounces much harder than desktop, which usually points at layout, tap targets or load time rather than at the writing.
- A page that is meant to produce enquiries has both a high bounce rate and almost no key events, which is the one combination that is unambiguously bad.
Everything outside that list is noise, and a report that leads with bounce rate every month is usually a report with nothing better to say.
Want to know why a page is losing people?
We will look at the page, the tracking behind it and what people do on it, and tell you which of the three is the problem.
Straight answers.
What is a good bounce rate?
There is no credible universal benchmark, and anyone quoting one is comparing pages that do unrelated jobs. A page whose purpose is to give an address will bounce heavily and be doing its job perfectly.
Judge a page against its own history and against pages with the same purpose, and judge it mainly on whether it produces the action you wanted.
Why did my bounce rate change when I moved to GA4?
Because the definition changed. Universal Analytics counted a bounce as a single interaction session. GA4 counts a bounce as a session that was not engaged, and engagement includes simply staying longer than ten seconds.
Most sites saw the reported number fall sharply. Nothing about visitor behaviour changed on the day of the migration.
Is bounce rate the same as exit rate?
No. A bounce describes the whole session: someone arrived and did nothing that counted as engagement. An exit simply records the last page somebody saw before leaving, and every session has one.
A checkout confirmation page with a high exit rate is a business working normally.
Does a high bounce rate hurt my Google rankings?
There is no evidence that Google reads your analytics account, and Google has said it does not use that data for ranking. Improve the page because it should answer the question well, not because of a number in a dashboard.
Where do I find bounce rate in GA4?
It is not shown by default. Add it as a metric when customising a standard report, or build a free-form exploration in Explore and select it there. Engagement rate is displayed more readily and carries exactly the same information.
Can I make my bounce rate look better without improving anything?
Yes, and that is precisely why it is a weak metric. Firing an extra event on load, raising the engagement timer or installing a tag twice will all flatter the number without a single visitor having a better experience.
If a report gets noticeably better with no change to the site, assume a measurement change until you have proved otherwise.
Where this comes from.
Primary documentation and published research behind the guidance on this page.
- Google Analytics Help: Engagement rate and bounce rate in GA4 (opens in a new tab)GA4 defines bounce rate as the inverse of engagement rate, not as a single-page visit.
- Google Analytics Help: Introducing the next generation of Google Analytics (opens in a new tab)
- Google web.dev: Web Vitals (opens in a new tab)The current thresholds for LCP, INP and CLS, and the 75th percentile rule.
- Google Search Central: Creating helpful, reliable, people-first content (opens in a new tab)Google's published self-assessment questions for content quality and experience.
Talk to the team
A short call, a look at how the business currently shows up, and a straight answer on what we would do first.