Rich-text Reply

Best Practices for AB Testing Engagement

ChrisSanfilippo 04-18-14

Best Practices for AB Testing Engagement

Here's what I'm doing and where I need help,


I start in Google Analytics and find the highest trafficked pages with bounce rates between 75-100%. I sort by page views highest to lowest, and start to identify the best pages to improve.


Now I want to use Optimizely to AB test to REDUCE the bounce rate. What's the best goal to use in optmizely to accomplish this? Is there a bounce goal? I didn't find the answer on either of these pages:


Thank you for creating this community!


Re: Best Practices for AB Testing Engagement

I think the best thing for you to do is integrate your Optimizely and Google Analytics accounts. You can view segments in GA and easily see their bounce rates.
-Bryant Jaquez
cofounder, Noble Creative.

brianlang 04-18-14

Re: Best Practices for AB Testing Engagement

Quick tip on Google Analytics - Use the weighted sort feature on Bounce Rate - it will show you where you can make the biggest improvements (based on page traffic and bounce rate) - it's magic Smiley Happy

As BRYANTJAQUEZ noted, integrate Optimizely with GA through custom variables to track bounce rate by variation. It might be better to choose a success metric that aligns more closely to your end business metric (such as a lead form sign-up, checkout, etc) - by improving a micro-conversion such as bounce rate, you might simply be shifting abandonment to another step further down the funnel, and not necessarily improving your bottom line.
Brian Lang

Re: Best Practices for AB Testing Engagement

[ Edited ]

Very good stuff! Thank you for sharing! 

"It might be better to choose a success metric that aligns more closely to your end business metric"


I think I may be using optimizely in a fashion normally not intended. This is for SEO!


I want to AB test pages to improve their bounce rates. The idea is that a lower bounce rate = better serp rankings. Since I'll be running this test on my pages that rank between 4-10 in the search results, it could be something that directly improves revenue due to the boost in highly valuable organic traffic.


I think some "ok" metrics to use a goal for the experiment would be time on page (x>5s), scroll depth (1-100), or just good old bounce rate.

kylerush 04-18-14

Re: Best Practices for AB Testing Engagement

In addition to integrating with GA, you can create complex goals in Optimizely that track a visit to any other page on your website.

Suppose the URL you are optimizing for bounce rate is This regular expression will match any URL on the website except for /pricing: https\:\/\/www\.optimizely\.com(?!\/pricing)

The regex is a negative lookahead explained in more detail here:

This regex goal will tell you if the variation is better than the control at getting a visitor to another page on the site. It's almost the same thing as bounce rate, but without the time aspect.

Re: Best Practices for AB Testing Engagement

I do apreciate that tip, it is not the ideal solution however. 

Bounce rate is already a poor measurement of engagement, so this would not be helpful in this particular situation. It may be helpful in another scenario, or for others so for that I thank you! It is a good tip to know.


I use a lot of event tracking in Google Analytics. This helps me exclude visits from the bounce rate when visitors watch a video, flip through an image gallery, or complete any other task that does not generate a new url. If only optimizley could use the Google Analytics "rules" for bounce rates for each page, taking GA events into consideration.

kylerush 04-20-14

Re: Best Practices for AB Testing Engagement

Have you checked out the Engagement goal? This is added to every experiment by default. It tracks visitors that click anywhere on the page. Engagement is essentially the opposite of bounce rate so you could use this goal to reduce your bounce rate.

Re: Best Practices for AB Testing Engagement

Hi Kyle,

I'm familiar with the engagement goal and to be honesy I do not like it.

Someone clicking on a page is not a good measurement of engagement.

If I had a video with a play button, or an image gallery, that might work.

But it will be different for every page and using a click universally as a goal means that the results will not provide very good insights in most cases.


Correct me if I'm wrong, or if I'm overthinking something, but just because someone clicks on a page doesn't mean they're engaged.

That's why using GA events would be great. I customize the bounce rate for EVERY page of my website, because that's the way you get the best measurement of how people are interacting with your site.

kylerush 04-20-14

Re: Best Practices for AB Testing Engagement

If you don't like the engagement goal to measure engagement then you can always use a custom event. This allows you to track engagement however you've defined it. On the Optimizely editor, I use a custom event to track engagement. If the user interacts with any feature in the editor I send out a custom event called "level 1 engagement". This tells me if the user is actually using a feature and not just clicking randomly somewhere on the page. This is nearly identical to what you're doing with custom events with Google Analytics.

More on custom events here:

Re: Best Practices for AB Testing Engagement

Hey Chris, as a fellow SEO guy I'm interested to know more about this test and what results you see. What type of content is this? What types of tests do you plan to run to improve engagement? This is something I've dealt with quite a bit, maybe we can share ideas.