How do I split test my website? – A/B Split Testing

written by William Patton on November 28, 2012 in Analytics with no comments

What is split testing?

Split testing is the process of compairing something against something else. In terms of website split testing it means your creating multiple versions of the same page and compairing them against each other to find out the best variant. When you have the results of the testing and can safely determine the best page and then use just that one. Or create another variant and test that against the winner of the pervious test. Continuous improvement is the aim of the game.

Isn’t split testing complicated?

Split testing can range from a simple color change, title change or minor rewrites all the way to an entire rearangement of content and design elements – it can be as simple or as compicated as you like.

A/B Split testing involves the creation of 2 versions of a page and randomly serving one or the other to different visitors and analyzing the data to find which version performs better. In the beginning, in fact all the time, I suggest changing only one simple thing on each page. Changing just one thing makes it easier to determine what has changed on the page that has affected visitor behaviour.

Why should I split test my website?

I shouldn’t need to tell you why you should be split testing your site. It’s about one main reason: improvement.

You should always be trying to improve your site one way or another and split testing is one of the best ways to do that. Have you ever been writing an article and you can’t decide what call to action is best? Or don’t know if it’s best to put that image full width above the introduction or a thumbnail at the side? Then test it and find out.

Don’t guess which is best anymore – test it!

There is a huge number of things you could test but I suggest you focus on things like page titles and descriptions (these are the things that searchers will see so make them count), image and ad placement, link color, font color/size/type and your CTAs. Change just one thing on each variant and let it run. Leave it to it’s own thing for a while untill enough data is collected to decide which variant is better. The amount of time you need to wait depends on the amount of traffic the page recieves. High traffic sites can probably make fairly good decisions based on only hours or even minutes. Medium sites, like my gaming site, take days-weeks to make an informed decision. Small sites might take months to collect enough data to see which variant is best.

I don’t get much traffic to my site, how can I split test?

If you run a site that gets little-to-no traffic then you will be waiting for a very long time to benefit from the data from your split testing. I would suggest that instead of doing split testing you should look into another branch of site improvement or SEO. Learn to write better content or do some keyword research and then you should split test it.

This post has been part of my Make Better Use of Analytics serise.