AB testing webpage CSS

AB testing is one of the most talked optimisation strategies for webpages. Google made it very popular by optimising it thoroughly in their pages. It’s using crowd-source intelligence for website optimisation…

The basic idea of AB testing is that you have 2 alternative pages that you want to test and see which of the two performs the best. The notion of performance can be which of the two alternatives led to more sells, or more clicks, or what ever you want to convert to.

Then the analysis is just a matter of computing the statistical relevance of the difference of the two alternatives (a basic two-sample proportion test, or a two-sample Z-test, for those who want the maths).

Google provides a great tool to do Website Optimisation with AB testing (and Multi-variable testing), but the problem is that you can’t really do CSS testing with it (there are some ways around it but I don’t like them). For this I decided to roll out my own AB testing for CSS alone.

It uses Javascript to apply styles to the webpage as it loads and it stores cookies that will track information for the conversion part. My basic research question for my experiments with sixhat.net is “How do CSS styles affect people website usability?”. This is a simple question with lot’s of sub questions associated with it.

Here I’m using sixhat.net as my lab. It’s simply my blog and probably you have seen some natural changes over time, if you are a regular reader (that I appreciate). But some of the answers I get from my toy website are then translated to bigger projects (for example ASSYST received some inputs from the AB testing done here.)

If you come here sometimes and find that the website really looks strange, maybe its is because I’m testing something weird. Don’t worry, each test takes little time and everything will be back to “normal” in a couple of days (usually)!