This week I got an email from Google, saying that I had been accepted to take part in the Beta test of the website optimizer tool. Sad as it may seem, I had eagerly been awaiting this invitation. Having signed up for it back in December, it has been a long time coming. So what exactly is this new tool and how will it help me? Basically, it is Google’s answer to A/B split testing software to be used primarily in conjunction with Adwords, but I don’t see any reason why it also wouldn’t also work for testing pages used in an organic SEO campaign.

For those of you who aren’t familiar with the concept of A/B split (or multivariate) testing it basically involves the scientific testing of various parts of a web page such as images, headings, layout, buttons, navigation to work out which versions your users prefer, or better still, which one is most likely to result in a conversion. A/B testing is the simple test of changing just one element of a page against another to see which works best, whereas multivariate is the equivalent of thousands of A/B split tests. It’s like post click marketing, helping you make the most ($$$) of the traffic that PPC and Search has driven to your site.

There have always been plenty of pieces of software out there to take care of this for you, but none that I am aware of that fit in nicely with Google Adwords and that are free! Welcome Google Website Optimizer!

So lets get started. I am running a PPC campaign on www.commercetuned.co.uk, so I will use this as my test subject.

The first thing that I need to do is decide which landing page I am going to use for the test. I have been directing traffic from the PPC campaign to an article on social bookmaking that I wrote, so I will use that page. The aim of the page is to educate users and to induce them to request more information about this topic. So how do I work out how many parts of this page to change? The Optimizer tool tells me that I can test up to 1000 variations in one go, but the more variations I have the longer it will take to get meaningful results. In fact the length of an experiment is determined by 5 key factors:

1). Number of combinations
2). Conversation rate
3). Website traffic
4). Estimated conversion rate lift
5). Percentage of participation
These points should all be fairly self explanatory - i.e. the higher the number of combinations the longer the test duration; the lower the conversion rate the longer the duration etc.

I am now going away to work on the two versions of my page, so will let you know how I get on!

Diana Gray - Search engine consultant, Commercetuned.