Taking Off the Rose Colored Glasses with Conversion Testing


“Many of these theories have been killed off only when some decisive experience exposed their incorrectness… Thus the yeoman work in any science… Is done by the experimentalist, who must keep the theoreticians honest.” - Michio Kaku, theoretical physicist and co-creator of String Field Theory, Hyperspace


In the creation phase, we often get so wrapped up in our work that we become part of our project and it becomes too personal to look at what we’re creating from an objective point of view. 

When launching new websites, this couldn’t ring more true. 

With all of the preparation, research, designing and redesigning, the ultimate test comes when the site is launched publicly and the user conversion rate reports come back in. 

Manual user testing of your website is a great place to start, and should be done with at least a couple of subjects. 

Usability expert Steve Krug notes that “testing one user is 100 percent better than testing none” and “nothing beats a live audience reaction”.  Beyond the initial testing, however, there is a need for mass scale feedback about the real ability of your site make a buyer out of the visitor. 

Your New Favorite Tool

Enter Google Optimize – the free testing tool for your site. 

Google Optimize is a powerful tool to test actual user reactions to your site, serving real time variations of anything on your site you want to test.  There are two ways of doing this - A/B split testing and multivariate testing. CXL have put together a great beginner's guide to Google Optimize.

Google has gone through a few different confusing iterations and names for their A/B testing tool. What used to be Google Website Optimizer (GWO) back in the day is now Google Optimize. Seems Google can't exactly decide if they want to serve this market, but let's hope it's here to stay. Given the prevalence of Google Analytics, AdWords, and Tag Manager, it just makes sense to have an optimization solution wrapped into all of it.

A/B Split testing tests different versions of just one element on your website – either a full page or just one area of the page.  For example, you can test the sexy, sleek version of your home page against the elegant and tasteful version.  Additionally, you can test whether your adult diapers website product page should have a picture your CEO demonstrating their use, or if a picture of an elderly couple walking their dog in the park would have a more positive effect on conversion rates. 

Multivariate testing is much more intense and is better for websites that get large amounts of traffic, as you need a larger sample size than A/B Split Testing.  This test allows you to test multiple variables on a page and is especially helpful because the variables most often influence each other.  If your cat clothing website needs to test its special Christmas sweater page, you can test different versions of your page navigation, the product description, and buy now button simultaneously to find the combination with the best purchase conversion rates. 

A/B Split Testing is best used for:

  • Low traffic pages
  • Low conversion rate pages
  • Very different concepts

Multivariate Testing is best suited for:

  • Large sample sizes
  • Variables that may interact with one another
  • Refining page elements

The Flow of Testing

As you test more and more, you'll get a feel for what works and what doesn't, in greater accuracy.

This graphic from Lynca shows the evolution of testing in separate phases:


Img source: Lynda.com


The key takeaway is that you should test the big things first to find the low hanging fruit, and then refine over time in terms of quality and specific things you're testing.

If you're grand new to testing and your landing page is garbage, changing the color from green to red on a CTA button will do very little compared to testing a complete overhaul of the page.


Tying in Testing with Analytics

Your A/B testing in Google Optimize should be based on actual data, most likely from Google Analytics. 

You can't blindly choose what pages to test based on pure gut feeling, you need some data, and that's where Google Analytics comes in.

If you're new to Google Analytics check out this guide with a list of 17 places to learn Google Analytics, or the absolute beginners guide from Moz.

When I was first learning I was overwhelemed with features and things to do, but take the 80/20 approach: focus on the most important things first.


Best Pages to Test

Knowing what pages to test and creating the variations of the elements you are testing is the hardest part of the process.  Here are the best pages to test:

  • Loser Pages:
    • Pages with high bounce rates typically over 50%.
    • Exit pages that are not meant to be – product page, about page, home page.
    • Pages showing little to no conversions
    • Pages showing conversions more than 25% worse than typical pages on your site for that page type
  • Popular Pages: Find the top landing pages on your site
    • Test these against a new version of the page to see if you can get even better conversions
  • High Value Conversions: Focus on the pages closest to a revenue source such as a purchase or download.



Apply the 80/20 Principle

Remember the 80/20 principle that states 80% of anything - conversions, traffic, sales - will come from just 20% of your site pages. This is a rule of thumb and may manifest in different ways - sometimes it's the 90/20 principle - but the basic idea is that you should focus on the top pages and content first.

This will allow you to get your early wins.


Website testing is a continuous process that should be refined continuously to create the best experience for users and to increase that golden conversion rate.  Happy testing!



Related Posts



No comments made yet. Be the first to submit a comment
Already Registered? Login Here
Sunday, 15 December 2019
If you'd like to register, please fill in the username, password and name fields.

Member Login

Business Insights & Tips


Jill Holtz
2343 Points
Tena Glaser
1395 Points
Michael Lane
802 Points
Ron Immink
732 Points
Fionan Murray
721 Points
View Leaderboard