Bounce rate vs engagement rate in GA4: what changed and what a good number is
Few metrics cause as much confusion as bounce rate. In Universal Analytics it meant one thing, in GA4 it means something else, and privacy-friendly tools use yet another definition. Here's what each one counts, and how to use the number without fooling yourself.
The old definition: single-page sessions
In Universal Analytics, a bounce was a session with exactly one interaction hit, usually a single pageview. Someone who read a 3,000-word article for ten minutes and then left counted as a bounce. That's why blogs and documentation sites routinely showed bounce rates of 70–90% even when readers were happy.
The GA4 definition: sessions that weren't engaged
GA4 introduced the engaged session. A session is engaged if any of these is true:
- it lasted longer than 10 seconds (adjustable up to 60 seconds in the data stream settings);
- it had a key event (formerly called a conversion);
- it had two or more pageviews or screen views.
Engagement rate is engaged sessions divided by all sessions, and GA4's bounce rate is simply its inverse: 100% minus engagement rate. That reader who spent ten minutes on one article is now engaged, not a bounce.
So when a site moved from Universal Analytics to GA4, its bounce rate usually dropped sharply, without any change in how people actually behaved.
Other tools, other definitions
Many lightweight and privacy-friendly analytics tools use the classic definition: a session with a single pageview. Some exclude sessions with custom events. Comparing a bounce rate from one tool with another, or with a "benchmark" from a blog post, compares different things.
Two practical consequences:
- Pages that aren't tracked break the number. If the second page a visitor opens doesn't have your analytics script (a blog on a subdomain, docs on a separate app), the session looks like a one-page visit and counts as a bounce.
- Single-page apps need route tracking. If client-side navigation isn't reported as pageviews, every SPA visit looks like a bounce.
What is a "good" bounce rate?
There isn't a universal one. It depends on the page's job:
- A blog post or docs page that answers a question can be a success at a high bounce rate. Look at time on page or scroll depth instead.
- A landing page whose job is to send people to sign-up or pricing should have a low bounce rate. A high one means the message or the traffic is off.
- A pricing page is best judged by clicks to checkout, not by bounces.
Compare a page with itself over time, and the same page across traffic sources. If visitors from one campaign bounce far more than everyone else, the ad promises something the page doesn't deliver.
Better questions than "is my bounce rate OK?"
- Entry pages: which pages do people land on, and where do they go next?
- Funnels: of the people who reach pricing, how many start a trial?
- By source: which channels bring visitors who stay, sign up and pay?
PageLens shows entry bounce rate per page next to time on page, exit pages and source, so you can see why a page bounces rather than only that it does. It's cookieless, so the numbers include visitors who would decline a GA4 consent banner.