Friday, 27 April 2012

Optimizely brings A/B testing into the mainstream

As I've mentioned before, one of the keys to success in an increasingly competitive world is having the right data - and lots of it (see More data usually beats better algorithms). Businesses (and politicians) are increasingly looking beyond traditional research (with all its pitfalls) to real behaviour. And as more and more customer interactions move online, the possibilities for collecting real behavioural data are expanding rapidly.

One method of collecting real behavioural data is "A/B testing". Simply put, A/B testing works as follows:
  1. creating multiple versions or variations of pages on your web site with subtle differences in design or messaging,
  2. serve the multiple versions concurrently to different users at random for period of time, (typically, in a mature site, only a statistically significant percentage of users would see the variations under consideration with the vast majority still seeing the tried and tested pages)
  3. track the results in terms of which versions or variations produce higher levels of the desired behaviour,
  4. discard the versions that provide lower levels of desired behaviour and keep the versions or variations that provide higher levels,
  5. and finally, repeat the process as often as you like (apparently Google ran some 7,000. A/B tests in 2011).
The variations could be simple design changes: are users more likely to click on a button when it is red or when it is green? Does increasing the size of the button,or bolding a key word increase the number of clicks? However, more substantial variations are possible, ranging from different copy to different combinations of products and features.

Although the origins of A/B testing can be traced back to Google as far back as 2000, it really only began to capture people's imagination when US President Obama began to use it in his 2008 election campaign. In this CNN clip, Mario Armstrong describes how A/B testing is increasingly being used in US politics.


Unfortunately, A/B testing has historically been difficult to implement. That is, until Dan Siroker, the ex-Googler who introduced Obama's campaign team to A/B testing, teamed up with another ex-Googler, Pete Koomen, to launch Optimizely. With it's intuitive graphical user interface, Optimizely allows non-technical users to develop A/B testing campaigns.

You can see a quick demo of how easy it is to use here:


As A/B testing goes mainstream, I am sure we can expect more tools in the market. One things is for sure:- the world of online marketing and web development has changed significantly and forever.