A/B Testing: What It Is & How to Do It


As marketers, we’re constantly making choices in our quest to create the most effective content possible. What color should this CTA be? Should I use humor in emails to my customers? Can this landing page headline be improved? The decisions go on and on, and unfortunately, it can be a bit tricky to determine if you made the right call.

An A/B test is a way to stop guessing and start knowing! 

An A/B test is one of my favorite tools. It’s the delivery of two different versions of a piece of content to randomly selected segments of an audience to collect quantifiable data about which achieved greater success toward a specific goal. 

A/B testing was the topic of a recent 'Inbound & Down' Question of the Day. You can check out the video with Morey Creative Studios Senior Strategist Danielle Esposito and Senior Web Designer Jeff Main here. And subscribe to the Question of the Day while you’re at it!

How do I run an A/B test?

Running an A/B test is relatively simple, but requires a few steps to be done correctly and yield helpful results. 
1. Decide what metric you’re measuring.

The purpose of an A/B test is to decide which out of two versions of a feature on your site was more successful. Therefore, the first step is to determine what constitutes success in the first place! 

Are you looking to improve the open rates of your emails? Volume of form submissions? These questions are not just important for your data analysis, but can also be necessary for using automated A/B testing tools, such as HubSpot’s, which can select a winner for you! 

2. Decide what it is you’d like to test.

Most A/B tests originate from a question. Normally, this takes the form of, “Would my audience prefer X or Y?” This can be about anything from colors to wording to images and more. The key is to limit the experiment to one element.

Think about it: If you ran an A/B test but included multiple differences between versions A and B, and one did better than the other, how would you be able to tell with confidence which factor was responsible for its success? Do yourself a favor and just pick one. Remember, you can always test another factor in the future if there’s more you’re curious about!

3. Build and deploy.

The specific tools you utilize to build your A/B test will depend on the platform you use to create your emails, pages, and other elements of your website. HubSpot, for example, makes building and running A/B tests for emails, website pages, landing pages, and CTAs easy! 

Whatever platform you build them on, A/B tests must be deployed properly to provide meaningful data. This includes:

  • Randomly selecting segments of the list or audience. 
  • Making sure that both variations of the test are being sent or viewed at the same time. 
  • Setting an appropriate amount of time. An A/B test of an email can yield useful data in a matter of hours, while tests for CTAs, on the other hand, might require months. 

4. Analyze the results and determine the winner.

This is perhaps the most exciting part of the A/B testing process, especially if you had a favorite version you thought would win. If you’ve followed the proper steps in developing the A/B test, this is also the easiest! It can be as simple as comparing a single metric, such as open rate. HubSpot can even select the winner for you! 

5. Implement the results.

Running an A/B test can be fascinating, and even a little exciting! What matters most, though, is what you do with that information once the test is over. 

Ask yourself: “What did we learn from this?” If you ran your test properly, it should give you a strong idea of your audience’s preferences, even if it’s just for one small factor. Now, you can capitalize on that new understanding by incorporating it elsewhere on your site or in your marketing activities. 

Don’t stop there! Just because you’ve performed one A/B test between two different variations, don’t assume the winner of that test is the undisputed best version of the content. Don’t be afraid to run another test in the future against another variable to see if there’s an even better version out there!