Bounce rate and exit rate metrics

  • Updated

Relevant products:

  • Optimizely Web Experimentation

This topic describes how to:

  • Measure bounce rate and exit rate metrics for your Optimizely Web Experimentation experiments and campaigns
  • Distinguish among bounce and exit rates as measured in Optimizely Web Experimentation, Google Analytics, and Adobe Analytics
  • Resolve implementation discrepancies between Optimizely Web Experimentation and your analytics platform

Bounce rate and exit rate are metrics that let you measure how engaging your website’s pages are. In the context of experimentation, bounce rate and exit rate metrics let you see how your experiments affect engagement levels. Typically, bounce rate measures the effectiveness of landing pages, and exit rate measures the effectiveness of down-funnel pages.

Bounce rate and exit rate metrics are only available for Optimizely Web Experimentation projects. 

Bounce rate

The bounce rate metric is useful for measuring visitor engagement when visitors are first interacting with your website. For example, you may hypothesize that changing your landing pages to mirror the value proposition used in your ad campaigns will create a more compelling experience and drive visitors to engage further down your funnel. Measuring  an experiment’s impact on the bounce rate of your landing pages can help you validate or disprove this hypothesis.

The bounce rate metric uses the following definitions:

  • Total bounces – Total number of sessions where visitors who were exposed to a variation viewed only this page

  • Qualified sessions – Total number of sessions where visitors who were exposed to a variation started the session on this page

  • Bounce rate – Total bounces ÷ qualified sessions

Example with 5 sessions:

  • Session 1 – Page A > Page B > Page C > exit

  • Session 2 – Page A > Page B > exit

  • Session 3 – Page A > exit

  • Session 4 – Page B > Page C > exit

  • Session 5 – Page B > Page A > exit

Using these definitions, Page A had 1 total bounce (in session 3) and 3 qualified sessions (sessions 1, 2, and 3). Therefore, Page A has a bounce rate of 33%.

Global bounce rate

Some customers want to track the impact of an experiment on their website’s global bounce rate: the percentage of visits where a visitor viewed only one page. To track this metric in Optimizely Web Experimentation, configure a global page (a page that fires everywhere the Optimizely Web Experimentation snippet is implemented). Then, configure a bounce rate metric on the global pageview event.
 
Knowing the global bounce rate is useful when your experiment makes changes to an element that appears on all pages with the goal of making your site more engaging (thereby reducing global bounce rate). There is no equivalent “global” metric for exit rate, which only has meaning in the context of a specific page.

Bounce rate in Optimizely Web Experimentation compared to Google Analytics and Adobe Analytics

Not all implementations of Google Analytics (GA) or Adobe Analytics (AA) will track interaction events. If your GA or AA implementation only tracks pageviews, then your GA or AA bounce rate calculations will closely match your Optimizely Web Experimentation bounce rate. If you have implemented interaction events, this section explains why your GA or AA bounce rate may differ from your Optimizely Web Experimentation bounce rate.

Google Analytics and Adobe Analytics both count a bounce on a page under the following conditions:

  • The visitor only views one page during their session

  • The visitor does not trigger any additional “interaction events” after viewing the page

By contrast, Optimizely Web Experimentation does not take into account subsequent “interaction events” when defining bounces. This means that in most cases, Optimizely Web Experimentation bounce rates will be higher than bounce rates in GA and AA.

Consider the following example session:

  1. Visitor begins a session on the homepage

  2. Visitor plays a video, triggering an “interaction event” in GA or AA and a click event in Optimizely Web Experimentation

  3. Visitor ends their session without navigating to another page

Optimizely Web Experimentation would count this as a bounce from the homepage, whereas GA and AA would not count this as a bounce.

Exit rate

The exit rate metric is useful for measuring visitor engagement on certain pages that are further down in your conversion funnel. For example, you may observe that your plans and pricing page has a high exit rate. Your hypothesis might be that adding a global navigation bar will give visitors more options to continue researching your product, and in turn you will see a decrease in the exit rate for your plans and pricing page.

The exit rate metric uses the following definitions:

  • Total exits – Total number of sessions where visitors who were exposed to a variation ended the session on this page

  • Qualified sessions – Total number of sessions where visitors who were exposed to a variation viewed this page

  • Exit rate – Total exits ÷ qualified sessions

Example with 5 sessions:

  • Session 1 – Page A > Page B > Page C > exit

  • Session 2 – Page A > Page B > exit

  • Session 3 – Page A > exit

  • Session 4 – Page B > Page C > exit

  • Session 5 – Page B > Page A > exit

Using the above definitions, Page A has 2 total exits (sessions 3 and 5) and 4 qualified sessions (sessions 1, 2, 3, and 5). Therefore, Page A has an exit rate of 50%.

Data freshness and bounce and exit rates

For Optimizely Web Experimentation to count a visitor as “bounced” or “exited,” we need to wait for their session to end (in other words, wait until the visitor has not triggered any events for more than 30 minutes). This means that the numbers displayed on the Results page for bounce and exit rate metrics will be at least 30 minutes behind other “real-time” metrics like binary conversions and revenue.

Compatibility with single-page applications

Optimizely Web Experimentation relies on client_activation events to determine when visitors move to a new “page” on your website. This information is necessary to accurately calculate bounce rate and exit rate. client_activation events are generated when the Optimizely Web Experimentation snippet activates, which typically happens when a page loads.
 
If you are implementing Optimizely Web Experimentation on a single-page application (SPA), you may have chosen to implement a call to Optimizely Web Experimentation's Activate API on each navigation. Calling this API causes the snippet to generate a client_activation event (among other things). Therefore, bounce and exit rate metrics will work as expected with SPAs that are set up in this way.
 
Currently, SPAs that trigger Pages via manual or conditional activation are not compatible with bounce and exit rate metrics.  

Similarly, the new Support for Dynamic Websites feature provides additional activation triggers "When the DOM Changes" and "When the URL Changes".  Sites that use these are also not compatible with bounce and exit rate metrics.

Implementation discrepancies

If you compare Optimizely Web Experimentation's bounce rate or exit rate to a similar metric in another analytics platform, you may notice unintuitive discrepancies. If you have confirmed that Optimizely Web Experimentation and your analytics platform define bounce and exit rate metrics in the same way, the discrepancy is likely due to inconsistent implementations.

Imagine a website where Optimizely Web Experimentation is implemented on marketing pages, but not post-login application pages. By contrast, an analytics platform is implemented both on marketing pages and application pages. Consider the following session, which consists of four pageviews:

  • Pageview 1 – landing page

  • Pageview 2 – homepage

  • Pageview 3 – login page

  • Pageview 4 – account page

If Optimizely Web Experimentation is implemented on all pages except the account page, we would calculate that an exit occurred on the login page (because we would not track the pageview on the account page). Conversely, an analytics platform that is implemented on all pages would calculate that an exit occurred on the account page.

To eliminate this implementation discrepancy, confirm that the Optimizely Web Experimentation snippet is implemented on all pages where your analytics platform is tracking events.

See also how to create bounce rate and exit rate metrics and add them to your experiments.