Bounce rate has the worst reputation of any metric and produces the most confusion. It appears in every report, gets mentioned in every meeting, and in most cases the conclusion drawn from it is exactly backwards.
The core of the misunderstanding: a high bounce rate is not a verdict on the page. It is an observation that only means something once you know what the page was supposed to do.
What GA4 actually counts as a bounce
In the previous version of Google Analytics, a bounce was a session with exactly one page view. Whether somebody read for four minutes or closed the tab after a second made no difference.
GA4 turned the definition around. It defines an engaged session first, which happens when at least one of these is true:
- the session lasts longer than ten seconds with the tab in the foreground,
- there are at least two page views,
- or at least one conversion event fires.
Bounce rate is then simply the inverse: one hundred percent minus engagement rate. That is a clear improvement, but it also means the figure depends on an arbitrary threshold. Ten seconds is a decision somebody made, not a law of nature. The threshold can even be changed in the settings, which quietly makes comparisons between two accounts worthless.
The page that is allowed to bounce at 89 percent
A dental practice had an 89 percent bounce rate on its opening hours and directions page. The agency report had it flagged in red.
What actually happened: somebody searched "dentist open saturday", clicked, saw the table, learned what they needed and left to get in the car. Four seconds, one page view, no further event. A bounce by every definition, and at the same time precisely the outcome the practice wanted.
The same 89 percent on a category page in a shop would be a disaster. The difference is not in the number. It is that one page does its job in a single glance and the other only does its job if somebody clicks onward.
So sort your pages into two piles once. Pages that answer a question conclusively. And pages meant to lead somewhere next. Only in the second pile is a high bounce rate an alarm.
The number that actually hurts is not in Analytics
The real warning sign is pogo-sticking: somebody clicks your result, is back in the search listing within seconds, and clicks the next one down. That is a completely different thing from a satisfied bounce, and it is the behavior you genuinely do not want.
It never shows up in Analytics, because your tool cannot see what happens behind the back button. You can corner it fairly reliably, though, by putting three figures next to each other:
- average engagement time for the page from GA4,
- the type of query bringing people in, from Search Console,
- and scroll depth.
A page whose top query demands an explanation, which is abandoned after eight seconds on average, and where almost nobody scrolls past the first screen, is a pogo-sticking case. The same eight seconds on a page holding a phone number is a success. Identical number, opposite meaning.
Why benchmarks rarely help
"The average bounce rate is 47 percent." Sentences like that appear in a lot of articles and they are close to worthless, because they average across page types. Blog posts, checkouts, contact pages and home pages folded into one figure.
Two comparisons are useful instead: the page against itself over time, and the page against other pages of the same type on your own site. If one out of eight product pages sits thirty points above the other seven, you have found something. If your whole site sits three points above an industry figure from somebody else's article, you have found nothing.
Bounce rate is not exit rate
Two metrics that get confused constantly, and the confusion leads to wrong conclusions. Bounce rate counts sessions that started on a page and ended there with no further action. Exit rate counts which page sessions ended on, no matter how many pages came before.
An example makes the difference clear. Somebody arrives on your home page, clicks through three product pages, and leaves the site on the third. For bounce rate this session is not a bounce, because there were several page views. For the exit rate of that third product page it very much counts as an exit. Confuse the two and you read a high exit rate on the cart page as a bounce problem and optimize the wrong place. A high exit rate in the middle of a checkout is an alarm. A high exit rate on the thank-you page after the purchase is the ending you wanted.
What you can actually change
Suppose you have found a page where a high bounce rate really is a missed goal. Then three things move the number in the right direction, and all three come back to the question somebody arrived with: the top of the page answers the query literally, the next step you offer matches the stage the person is in, and the page loads fast enough on a phone that anyone reaches the content at all. Each of those fixes, with a concrete method, is in reducing bounce rate; the technical side of load time in improving Core Web Vitals.
What not to do: game the ten second threshold with an automatically firing event. It takes five minutes to implement, improves the number immediately and changes nothing about reality. All you have done is destroy the last metric that was still telling you something.
A test more honest than any metric
Take the top query for a page, type it into Google on your phone, click your own result and count the seconds until you can see the answer. If you had to scroll, you know where the bounce rate comes from.