What It Actually Means
The important thing to know before discussing it is that the definition changed. In older analytics a bounce was a single-page visit, however long they stayed. In GA4 it is the inverse of engagement, where an engaged session means more than ten seconds, a conversion, or two or more pages.
So a visitor who read a page thoroughly for four minutes and left counted as a bounce in the old tool and does not in the current one. Any comparison across that change is meaningless, and comparisons against published benchmarks are worse, because you rarely know which definition produced them.
Why High Is Often Fine
A bounce is not a failure. It frequently means the page worked.
Somebody searches for your opening hours, lands on your contact page, reads them, and leaves. That is a perfect outcome and it counts as a bounce. The same applies to anything answering a specific question: a definition, a phone number, an address, a price.
It applies to plenty of good content too. A visitor who reads an article to the end and leaves satisfied has had exactly the experience intended.
This is why a high bounce rate on its own justifies no action. The question is always whether the page’s job requires a further step, and for many pages it does not.
When It Genuinely Signals A Problem
The number becomes useful when it is unexpected for that page’s purpose.
A landing page built to generate enquiries with a high bounce rate is a real problem, because the page exists to produce an action.
A sudden change on a page that was previously fine. That usually means something broke, or the traffic mix changed, and both are worth knowing.
A large gap between mobile and desktop on the same page, which normally points at something specific: layout, load time, an intrusive element.
A gap between traffic sources. If one channel bounces far harder than another on the same page, the expectation being set before the click does not match what arrives.
That last one is the most commercially useful, and it is a common finding in paid search where an ad promises something the page does not immediately deliver.
What To Ask
- What is this page for? Decide whether a second interaction was ever wanted before treating the number as bad.
- Which definition is this, and has the tool changed? Necessary before comparing to anything historic.
- How does it split by source and device? The aggregate is not actionable. The splits usually are.
- Did it change, and when? A shift is informative in a way a level is not.
What To Use Instead
For pages that exist to produce an action, the direct measure of that action is better in every respect: enquiries, purchases, calls.
For content pages, engaged time and whether people continue into the site say more about whether the page worked. And where the concern is whether visitors are finding what they came for, the search terms bringing them and what they do next answer it far more directly than a single aggregate percentage ever will.
More terms are in the glossary.
