What is exit rate?

The metric always looks at one specific page, never at the whole site. Every session ends somewhere, so every session produces exactly one exit. What matters is which pages the exits pile up on and whether that ending was the intended one.

The usual confusion is with bounce. A bounce describes a session that starts and ends on the same page with no further activity. An exit describes the end of the session no matter how many pages came before. Every bounced session also produces an exit, but most exits are not bounces: they follow a journey that may have been long and worthwhile.

Google Analytics 4 adds a wrinkle, because bounce was redefined. It no longer means «only one page viewed» but the opposite of engagement rate: a session counts as a bounce when it clears none of the engaged-session thresholds, namely lasting longer than ten seconds, firing a key event, or collecting two or more views. The exit, by contrast, stayed a fact about sequence rather than a judgement about the quality of the visit.

That is why the two metrics answer different questions. Bounce asks whether the visit ever got going. Exit asks where the journey came to a stop.

The formula is simple: exits recorded on a page divided by views of that same page, times one hundred. A page with 400 views and 120 exits has an exit rate of 30 %.

The trouble sits in the numerator. Universal Analytics shipped exit rate as a standard metric in its content reports, right next to bounce rate. Google Analytics 4 does not include it. What it does offer is the Exits metric, defined as the number of times the last event of a session happened on a page or screen, together with its counterpart Entrances. Neither appears in the standard reports, so you have to go to Explore, build a free-form exploration, set page path as the dimension and add the Views and Exits metrics.

The percentage is worked out away from the interface. The usual routes are exporting the table to a spreadsheet, setting up a calculated field in a reporting tool, or requesting both metrics from the Data API and dividing them in your own script.

One quirk of the event model belongs here. Since GA4 counts events rather than pages, the last event of a session need not be a page view: it can be an outbound click, a download, or a scroll event. The exit is attributed to the page where that last event occurred, which is not always the page the user thought they were reading. On top of that comes the session timeout on inactivity, set to thirty minutes by default, which turns any long pause into an exit.

The decision that hangs on this metric is one of priority: which page you touch first. A purchase or contact journey is a chain of steps, and exit rate shows which link breaks. If the shipping-details step is where exits gather, the work belongs there and not on the product page.

The reading only works if you know the page type. On a thank-you page, an order confirmation, or an article that fully answers the query, a high figure is the expected outcome: the user got what they came for and left. On an intermediate page, the same figure points to friction, to a broken promise, or to the absence of a visible next step.

This is why no general benchmark exists. Any number presented as a «good exit rate» without saying which page type it refers to is noise. The useful comparison runs against the same page over time, before and after a change, or against pages of the same type within the same site.

Combined with the economic value of each step, the metric sorts the work: first the pages where many exits coincide with a lot of lost value.

• Group pages by type before comparing anything: forms, product pages, articles, confirmations. A table sorted by percentage without grouping mixes wanted endings with problematic ones.
• Multiply the number of exits by the value of the step instead of looking at the percentage alone. A page with 80 % exits and thirty views a month matters less than one with 35 % and fifteen thousand.
• Build the metric once in a saved exploration, with page path plus the Views and Exits metrics, and reuse it at every review so the figure stays comparable across periods.
• Cross exits with traffic source. If a page holds organic traffic and loses paid traffic, there is usually a match problem between the ad and the content.
• Split the data by device. Jumps on mobile often come from elements that cover the content, awkward forms, or load times that go unnoticed on desktop.
• Note the dates of template and content changes and check the series against them, rather than reading isolated weekly swings.

• Treating every exit as a failure. The order confirmation and the submitted contact page have to gather exits, because they are the ending you were aiming at.
• Copying a benchmark read somewhere else. Without page type, sector and measurement method, that number means nothing and leads to redesigning pages that worked.
• Confusing the metric with bounce and drawing conclusions about content quality from a figure that only describes the order of events.
• Forgetting the session timeout on inactivity. Long-read pages accumulate exits simply because the user stays with the text for more than thirty minutes.
• Comparing GA4 figures with the ones Universal Analytics used to give as if they were the same measurement. The event model, the session definition and the handling of consent all changed.