close

A/B Testing

Synonyms: split testing

A/B testing is a research which allows you to compare and analyze two or more versions of one element, page or user flow, and understand which of them can let you achieve a favourable result. 

 

Globally A/B testing isn't just about the web and it’s migrated from classical marketing. Here is one short curious lecture about this approach from the «Critical Thinking for the Information Age» course.

 

A/B testing will help you to decide which option will be easier to implement, based on facts and metrics. A/B testing is also helpful with explaining to the client why this design will work and the other won’t. This approach gives you an opportunity to increase website conversion as well, by bringing the best solution on the table.

 

You can test almost any content (web pages, ads, e-mails) or element on your page (headings, CTAs, buttons, images, texts and forms). The process takes no time  time but is worth the effort.

 

Data-driven design illustration by Marpipe Design

 

First step is to define a goal. What do you need to achieve? What is the purpose of your content: user engagement, conversions, clicks? Once you are set with a goal, you can proceed to hypothesis brainstorm. Don’t hesitate to create as much as you can so you will be able to choose the best. Based on your hypotheses you now have an opportunity to bring your design options to life — two or more — and test them. But don’t run ahead! After you have defined a goal, hypotheses and created your designs, you should think about targets. What is the criteria for the best solution? How many target actions do you need to get to decide that A-version or B-version is better? 

 

After figuring this out, you may determine the coverage of testing (how many people should visit the page to consider results as reliable?) and duration of experiment (at least a week as users behave differently on various days of the week).

 

There is another step of preparation which is optional. Уou can take the A/A testing to see if the data is valid. This means you will divide your audience and compare reactions on the same design and see if the performance is clear. Finally, after all the run-up you can provide your audience with designed options and start the A/B testing. Don’t forget to analyze results and metrics, identify which option brought more desired actions (purchases, registration, page openings, etc) and make a decision on what solution to use in your project.

 

There are plenty of resources that can help you to make the testing on your own if you don’t have any reliable tools: Optimiizely, AB Tasty, Unbounce,AB Test Guide, VWO, Google Optimize, Google Experiments, Changeagain and many more.

 

That was a brief description on what you should know about A/B testing. Don’t hesitate to use one of the research tools and feel free to dive in the world of UX! We are here for all the questions that might appear on your way.

share the term