UX for Beginners

Chapter 95. A/B Tests

User research and user psychology are great ways to predict how users will behave and what they will do. But we don’t just want to predict. We want to know.

Let me paint the picture a little bit first...

Imagine that you want to design a page to sell shoes. And, of course, you want to sell as many shoes as possible. What do you think would get more people to buy?

A video of the shoes?

Complete shipping details before they click “buy”?

The logo of the shoe brand?

A money-back guarantee?

How do you choose?

If your first thought was, “Ask the users!” that’s not a bad idea. But often—when the options are all subjective, like this—asking people just confirms that different people like different things.

So, how can you decide between subjective things, like a boss?!

Design all the things! Then, you launch all the options at the same time, as an A/B Test.

What Is an A/B Test?

An A/B test is a way of asking thousands or millions of real visitors which option is best.

The tests make sure that each unique visitor only sees one option. Then, after enough people have gone through your test, you can see which version of the design created more clicks. The test should also measure the “confidence level” statistically so you know when you’re done (don’t stop it too early!)

You can do it with 2 versions, or 20 versions. Just remember: only part of your traffic can see each one, so the more versions you are testing, the more traffic or time you will need.

Some tips

1. A/B Testing is usually free, ...