What bounce rate measures in each platform
In the old Universal Analytics, a bounce was a session in which the tracking code recorded exactly one interaction — one page, no event, no second hit. Because there was no second hit, the time spent could not be measured either, so a visitor who read a page carefully and left counted the same as one who arrived and closed the tab.
GA4 measures the opposite thing and then flips it. A session counts as engaged if it lasts longer than a configurable time threshold, or fires a key event, or includes more than one page view. Bounce rate in GA4 is simply the share of sessions that did not qualify — the inverse of engagement rate, and nothing else.
Elsewhere the word means something unrelated. In email marketing a bounce is a message the receiving server refused, split into hard and soft. Ad platforms generally do not report bounce at all; they report clicks against landing page views, which measures whether the page loaded rather than whether anyone read it.
Why the cross-platform difference matters
Bounce rate is one of the few metrics a business owner recognises without being taught, so it turns up in reports, proposals and arguments. When the underlying definition has changed but the label has not, people compare figures that were never measuring the same behaviour and reach confident conclusions that are simply wrong.
The stakes are practical. A migration from Universal Analytics to GA4 usually makes bounce rate look dramatically better, and nobody changed anything on the site. Read the wrong way, that becomes evidence for a redesign that worked, an agency that earned its fee, or a channel that improved.
Where cross-platform bounce rate comparisons go wrong
The most common error is holding a GA4 figure against a historical Universal Analytics figure, or against an industry benchmark published before GA4 existed. Neither comparison is valid, and no adjustment makes it valid.
The second is assuming a bounce is a failure. Plenty of pages exist to answer one question — opening hours, a price list, a phone number. In Nepal, where a large share of traffic arrives on a phone, the ideal outcome is often a single page view followed by a tap on the call button, which the browser records as no further engagement at all.
What to do about it
Report engagement rate rather than bounce rate, and state the definition in the report so nobody has to guess. Where a client insists on the familiar word, write next to it what GA4 actually counted.
Then measure the thing you really care about. Average engagement time tells you whether the page held attention, a key event tells you whether it produced the action, and both survive comparison over time. If you need help settling which numbers your reports should carry, that is a question of measurement setup rather than of page design.