Calculating Exits in Looker Studio from GA4 data - a new approach

GA4's Exits metric isn't available in Looker Studio. Here's how to create a calculated metric for Exits using your GA4 data.

Google Analytics (GA4) has a useful metric called "Exits" which is defined as "The number of times that the last event recorded for a session appeared on a particular screen". Sadly, this Exits metric is not available in Looker Studio when you connect your GA4 data using the official connector (much to some people's annoyance).

So is it possible to create an Exits metric directly in Looker Studio from GA4 data?

GA4's Exits metric - not available in Looker Studio

To answer that, we need to think about what happens when a user doesn't exit. If I enter directly on page A, then click to page B, two things happen: a View (page view event) with page A as the Page Location, and a View with page A as the Page Referrer. Like this:

However, if I enter and exit on page A, there will only be a View with page A as the Page Location.

So, in theory, if we take the number of Views where page A is the Page Location, and substract the number of Views where page A is the Page Referrer, we'll be left with the number of exits for page A.

Looking this up for individual pages in Looker Studio is kind of fiddly, so we can build a data set for this by blending our GA4 data with itself (Resource > Manage Blends in Looker Studio). The first table has the dimension Page Location and the second table has the dimension Page Referrer. Both tables have the metric Views:

Use a Left Outer join, and add a join condition that matches Page Location (Table 1) to Page Referrer (Table 2):

Then Insert a Table into your Looker Studio report using your new Blended Data data source. Add the dimension Page Location (from Table 1) and your two Views metrics:

That Views metric from Table 2 is actually Referrals (i.e. referrals from the page in question) so you can rename it to avoid confusion. Click on the word SUM to the left of the metric name in the Setup panel, then enter your new name:

With these two metrics, we can now calculate our Exits metric. Still in Setup, click on Add metric and then ADD CALCULATED FIELD.

Our formula for this new metric is simply:
Views (Table 1)-Views (Table 2)

The green tick will appear to show that the metric is valid:

Looking at the table data for a couple of pages on my own site, I can see that my homepage has far fewer exits than one of my popular guides:

We can go one step further and calculate an exit rate. Add another calculated metric, this time using the formula:
(Views (Table 1)-Views (Table 2))/Views (Table 1)

And set the Data Type to Numeric > Percent:

This even more clearly shows the difference between my homepage and that popular guide page:

Limitations of this approach

So is this a good direct replacement for GA4's Exits metric? Not really, because there are a few user scenarios that can skew the data. The most common one relates to the user opening internal links in a new tab.

Let's say the user is on page A. From there, they open page B and page C in new tabs. This gives us one View where Page A is the Page Location, but two Views where Page A is the Page Referrer!

In our table, we could end up with more Referrals than Views for a particular page, creating an exit rate of more than 100%. Clearly this doesn't make sense.

Page reloads can also cause oddities. Let's say a user lands on page A, reloads the page twice, then clicks through to page B (before exiting). This gives us three Views for A but only one Referral, resulting in an Exit rate of 67%. So a reload here is being treated as an exit, which again isn't really right.

Despite this, I think these calculated Exits and Exit Rate metrics in Looker Studio do have some value. Certainly they show on my site that individual guides have a much higher exit rate than the homepage or category pages, which is what I would expect to see. The metrics, despite their flaws, could be used for tracking changes over time or the impact of improvements.

And it's been an interesting experiment for me if nothing else!