Keyword segmentation can turn a failed test into a big success

Testing and optimization has really increased in popularity over the past couple of years.  With the increase in popularity has come an increase in what testers are using the software for.  Organizations, both big and small, are going above and beyond AB and MVT testing and incorporating more advanced tactics such as segmentation and personalization. 


For those not yet taking advantage of these advanced tactics, segmentation is the process of splitting/filtering your test data by different visitor attributes, whereas personalization is the process of delivering content that resonates well with specific types of visitors.  Segmentation is a great stepping stone towards personalization in that it helps you understand more about your specific audience(s) and generally what motivates them to take action.  In this article, we’ll focus our attention on segmentation specifically and how it can be used to personalize site content based on your segments.


One of the great things about segmentation during testing is that it is super easy. You can literally segment your traffic based on a plethora of visitor attributes including keyword, referring traffic, geography and many, many more.  More importantly, the benefits of segmentation are priceless. 


To prove my point, consider the data below taken from a recent test we ran with one of our clients.  The data provides us with two key insights into why it is important to segment your tests.


  • The first bar, representing ALL traffic combined, shows that for an unsegmented test, we would have observed a 5% decrease in conversion rates as a result of introducing the challenger and concluded that the control won. This wouldn’t necessarily be accurate considering point #2.
  • The remaining bars, representing a couple of keyword groups we used to segment the test data, shows very different results among the different keyword groups. This is an important insight because instead of saying this test was unsuccessful (if we had only looked at all traffic), we walk away from this test with a personalized experience for several keyword groups.  We could essentially show the challenger for the branded keyword visitors and the original experience for quote keyword and cost keyword groups.  A great win and learning!

Lift vs Control.JPG


As an extra added bonus, now that we understand the importance of segmentation, let’s dig into specific trends we’ve seen recently.  Over the past several months, we've noticed a common theme: branded visitors tend to prefer a streamlined experience, whereas non-branded visitors tend to need additional information to aid the decision making process.  Below are a few examples of this finding (Screenshots blurred to protect the innocent):


Example 1: Search landing page for Credit Reporting Business.


 Screenshot 1A.png

 Screenshot 1B.png

Branded Visitor Preference

Non-Branded Visitor Preference


Example 2: Search landing page for Home Phone Service Business.


 Screenshot 2A.png

 Screenshot 2A.png

Branded Visitor Preference

Non-Branded Visitor Preference


Key takeaway: When thinking about your next set of tests, consider adding segmentation as part of your strategy to get the best possible outcome and garner segment specific insights.  Segmentation will not only improve your chances of a win, it will help you personalize the experience based on your many audiences.


Do you use segmentation on your website? Let me know in the comments below.


Read more from the HP Optimost team:
When rubber meets the road for customer experience by Uri Kogan
E.ON and Choice Hotels deliver customer experience success with HP Software by Lindsay Sterrett
Are you in control of your web optimization program? by Uri Kogan
Got great test results during your sale period? Don’t expect them to last year round by Sarah Haggett

Search the HP Software Blog
Showing results for 
Search instead for 
Do you mean 
Follow Us
About the Author
Enjoy the ease & flexibility of HP Exstream Cloud Production.