3 Key Steps to Starting an A/B Test

Posted by Robert Brady - June 23, 2010 - Design/Development, SEM - 3 Comments

Ever wondered if your site would generate more sales/leads with a larger call to action (CTA)? Perhaps reducing the amount of text and using bullet lists instead? Even the experts will tell you “It depends…”, but with 3 easy steps you can let your customers tell you.

Step 1: Determine Your Goal

You need to clearly state what the goal of your test will be. Some people call this a research question, others call it a hypothesis. Most of the time is will sound something like “Which version of my homepage will lead to more sales/leads?” Just remember that each change you make should have this in mind.

In connection with determining your goal, make sure you have the ability to measure your success. If you’re using PPC, this is done with conversion tracking code on the thank you page. If you’re using Google Analytics, you’ll need to create a goal. Whether a sale or a lead, you need to be able to measure a successful action.

Engagement Goals:

Not only can you measure your success by total conversions or leads, but you can measure success by a visitors engagement.  Take into consideration the following:

  • Increased Time on Site
  • Decreased Bounce Rate
  • Increase in Pageviews
  • Increase in Social Sharing
  • Increase in RSS Subscribers
  • Increase in E-mail Subscribers
  • Increase in Video Views

Step 2: Design An Alternative Page

Perhaps you want to try a different headline. Maybe put a contact form in the sidebar. Try shortening your copy by 75%. And yes, perhaps even changing the color of the button(s). The key here is to make changes that are radical. Not just a little different, but a lot different.

For example, consider these two pages from a Marketing Experiments case study:

Here you can see a drastically different approach where the Treatment keeps the call to action above the fold and is more visually appealing. However, testing is vital because in this test the control outperformed the treatment by over 50%.

Once you’ve got the design created, post it to your website under a new URL like www.yoursite.com/test.html. Then move on to Step 3.

Step 3: Set up Google Website Optimizer

Go to Google.com/WebsiteOptimizer and sign in with your Google account info (or create an account, they’re free). You want to select the first option for an A/B experiment. You’ve already created the alternative, so check the box at the bottom and click next. Now you’ll just follow the instructions provided by Google (or email them to your website guy) to get your first test up and running.

For Detailed Instructions on setting up an A/B or Multivariate test check out Fundamentals of A/B and Multivariate Testing.

About the Author

Robert Brady

An Idaho farm boy working in the world of internet marketing, all my metaphors involve plants, cows or tractors. I've been managing PPC for 6 years and specialize in conversion rate optimization (CRO). Follow me on Twitter @robert_brady or email robert "at" righteousmarketing.com.

This is really an easy way to target your content to your visitors and customize the details of a page to best convert. It seems to be different for every page or subject but this is still a good source of info thanks.

Private Label Ebooks
Private Label Ebooks

In some cases I believe making radical changes to your page can cause more issues. For example when you want to test a new design or position for your opt-in subscriber box it is best to split test just these elements to determine which converts better rather then testing a completely new page and design.Once you determine the elements that convert well you incorporate them into the working design. The more split testing you perform on smaller elements of the page will surely result in a much better conversion rate for the purpose of the page.


  1. […] copy and changing the destination URLs. This is the simplest way, but you can get a more fancy with A/B testing. The key is to let your customer decide which landing page is better through voting with their […]