Split testing is awesome. A simple split test (also known as an A/B test) can be the easiest and cheapest way to test whether your creative content and layout ideas are going to work for your site or not. Now, Google is planning to make it even more effective with the beta launch of Google Content Experiments.
How to Manage Split Testing
If you’re familiar with split tests, you’ll know that Google has long offered a free tool for this purpose, Google Web Optimizer (GWO). Content testing in Google Analytics (GA) performs much the same function. The brand new Content Experiments resides entirely within the GA interface; the Google team maintain that they have rebuilt this tool from the ground up to be part of GA, and that GWO is set to be retired soon.
Split testing is a method of optimising your marketing efforts by trying out different versions of the same web page on separate groups of people. For example, you can create two versions of a web page – version A has a big flashing green button at the top, while version B has smaller pink button off to one side. You can measure which version leads to better results, and use that one in future. Here’s a great example of how removing a security icon increased a brand’s conversions by 400% –sometimes the things that work can be quite unpredictable.
In essence, split testing lets you fail fast! With a minimal investment (moving some buttons, finding new pictures, maybe a little design work), you can launch an alternate universe where your site looks different, and then measure how much better or worse your site performs.
There are some basics that need to be in place before you start your split test. First of all, you need to have a variation page to test with – design it, make your changes, and have it ready to launch. When you start the Content Experiment set up in GA, you will be provided with a special snippet of code to be added to the original version of your page.
If you are concerned about the risk of showing customers an unappealing visitor page, you can limit your exposure: only a certain percentage of visitors need be involved in the experiment at all – for example, if you have 100 000 visitors a month, you can choose to only experiment on 10 000 of these visitors. This will mean that, of the 10 000 visitors 5 000 will be served the original version, and 5 000 will be served the variation page.
Google Analytics Got This
One of the best features of the Content Experiments release is that Google Analytics will handle the split between your original and variation pages – this means you don’t have to worry about special coding to serve different pages to different visitors. Better still, Google Analytics will start to divert traffic away from a poorly performing page in the split test – if your variation page is a complete disaster, less and less traffic will be diverted to it. Likewise, if your variation page is a smash hit, more and more of the test traffic will be diverted there.
It is essential to make sure you pick a page that will generate enough traffic to give you statistically significant results quickly – there are plenty of free calculators online that will tell you how many visits you need in order to draw a statistically significant conclusion. There’s nothing worse than waiting months for an answer and not having a clear indication of what you should do.
Google Analytics will also set up cookies to make sure that the same pages are served to the same visitors every time they visit – it would be very confusing for a visitor to come to the site twice in one day and see two completely different formats.
The Content Experiments tool will declare a winner as soon as a critical sample size is reached. If no winner comes up, you were clearly not bold enough – see this article by Wired for best practices to guarantee that your split test works for you.