What’s the goal of A/B Testing?


What’s the goal of A/B Testing?

In a word: conversions. If you’re unfamiliar with the concept, “conversion” just refers to a visitor on your site doing what you hope they’ll do. That could mean purchasing a product, signing up for a service, or even just sharing something with their friends on social media. What a conversion means, in other words, is largely up to you. At any rate, A/B Testing is just a way of maximizing those conversions.

What does it entail?

First we try and zero in on where we think visitors are likely to convert. There are usually some pretty obvious places to take a look at first: purchase links, newsletter forms, etc. Sometimes, however, conversion optimization isn’t so straightforward. Maybe users are getting confused when they try to convert. Maybe the conversion process is too much of a hassle for some of them.

What then?

Then we experiment. We create multiple versions of the page, send some visitors to the original version and some to the new version, and pay attention to how they respond. If by simplifying a payment form we find that conversions increase, we change the whole site over to the new version.

Sounds simple enough.

It is simple in principle, but in practice it can get complicated pretty quickly. A/B Testing is when we compare just two versions of a page but there’s another concept called “Multivariate Testing.” It’s basically the same idea but with more variables. Let’s say you’ve got an asset on your site that consists of a graphic and some copy. We want to try two versions of each component and every combination therein to find the best combination. By adding just one extra variable to our analysis we’ve increased the number of versions we’d like to test from two to four. If we were to look at a third variable…well, you get the idea. The variations increase exponentially.

Okay, I changed my mind. That sounds complicated.

It can be but we’ve got a strong ally on our side: data. The more traffic your site tends to generate, the more data points we have and the deeper into Multivariate Testing we can go. There’s an optimal ratio between how much traffic your site has and what we can meaningfully test and we can help you find that ratio.

How do you do that?

Partially with a tool called Optimizely.

What’s that?

Glad you asked! Let’s look at that in the next article.