Thursday, January 20, 2022

Why Bounce Rate is a Terrible Benchmark

Are you measuring how your site or site section’s bounce rate compares to other sites? Don’t! Bounce rate is incredibly hard to standardize, not only across websites but also across your website, and acquisition channels. One of the major reasons for this issue is a poor understanding of bounce rate.

What is a bounce

The bounce and bounce rate definitions may be different depending on where you get your information. The Web Analytics Association defines a bounce as “a visit with one pageview.” Despite that being under a “Google Analytics Glossary” page, that definition is not aligned with how Google Analytics determines a bounce. Forget what anyone else, even your boss, says; go to your analytics provider to understand how it determines a bounce.

Google explains that in Google Analytics a bounce is calculated as a session that triggers only a single request to the Analytics server. When the page loads, a pageview request is sent to the server but requests other than pageviews, called interaction hits, can be sent from the same page, too. Imagine a user that visits a page that begins playing a video once the user has scrolled it into the frame. The user enters the site on that page, scrolls far enough for the video to start playing, scrolls past the video without watching, and exits the page. Is that a bounce? Depending on how your site is tracking the video event or any other on-page events, it may or may not be considered a bounce.

Events in Google Analytics default as “interaction hits.” When an interaction hit occurs in addition to the initial pageview, then the visit, or session, is not a bounce. If you are using events to track the video play, that event is likely kept to the default interaction hit, which means the user scrolling past it without watching and immediately leaving would NOT be a bounce. However, events can be set as “non-interaction hits” by setting the “non-interaction hit” parameter to “true.” Learn how to set the parameter to true in Google Tag Manager or with a hard-coded analytics.js or ga.js script.

How to properly benchmark bounce rate

Would you or your organization consider auto-played videos an interaction? What if competitors do and you don’t or vice versa? How might other pages on your website appear from an engagement standpoint if they don’t have automatic video plays bringing down the bounce count? Because sites and even pages within the same site can serve such different purposes, we shouldn’t be comparing bounces and bounce rate to other pages. CXL explains how two different sites (or sections on the same site)—a blog and an informational page—will have very different bounce rates because of how they serve the user’s intent. So what is bounce rate good for? You can use bounce rate as a metric to help improve the specific page.

Even when benchmarking the same page there are some caveats. Be sure events that aren’t triggered by a user interaction are not set to the default “interaction hit” to unnaturally set the bounce rate low. Some examples are auto-played videos, scroll tracking, and banner ad views. Toggling content, like an accordion or tab, is another one to set clear expectations for interaction hit, or not, and make sure that is consistent across the site. Another caveat to setting a standard or bounce rate goal for a single page is to consider how the acquisition channels might be affecting those bounces. From the CXL article again, a poor bounce rate from one acquisition channel might signal that the message is not providing the right context or is being presented to the wrong audience. A bad ad or ad targeting could raise the bounce rate making it seem as though the page is subpar.

In case you couldn’t tell, I don’t care much for bounces and bounce rate, and don’t include those in reporting. Happily, with GA4 we won’t even have to worry about it! Bounces and bounce rate do not exist in GA4 and instead are replaced with engaged sessions and engagement rate. I love Krista Seiden’s definition of engagement rate as an “inversion” of bounce rate, “measuring active interaction rather than the lack of it.” If you aren’t ready or just don’t like GA4 (yet), consider taking its engagement measurement as a way to better define engagement with the data available in Universal Analytics. Engaged sessions are those when the session

  • has lasted more than 10 seconds
  • resulted in a conversion event
  • had two or more page/screen views.

Some ways to analyze those in Universal Analytics are to look at the other engagement metrics, like average time on page when looking at page-level data or average session duration and goal completion/conversions when reviewing session-level data, like the Landing Pages report (which I’ll be talking about in my next post).

3 comments:

  1. Thanks for your post. I think your point about bounce rate not having much meaning without the context of page intent/uses to help define a win or a loss is an important one to remember.

    ReplyDelete
  2. I loved reading this because it gave me an entirely new perspective from what I wrote. As someone who is extremely new to web analytics, this helped me understand this tool even beyond what I read in class!

    ReplyDelete
  3. The definition of engagement rate as the opposite of bounce rate, and much more worth measuring, is fantastic. I can also see how bounce rate isn't a good metric to stick in a Big Quantitative Number Cruncher, but it can be good to analyze a specific page. Thanks for the detail on how bounces are defined and how you can customize them, especially for video!

    ReplyDelete