Genetic A/B Testing
I’m not sure I remember being this excited about a piece of open source code. While the premise is fairly simple, the algorithms at play and the potential uses are vast. Before I get too far ahead of myself, please let me clarify. I’m talking about a piece of software called Genetify.
The code, in a nutshell, allows for the real time testing of variations on aspects of a website. These aspects range from html elements to css styles to javascript variables. With a few small lines of javascript and a several variations in the code, you can let ‘genetics’ take care of the rest. For more information on how the software works, check out the usage section of the Github page.
I first read of this code being used on shoemoney.com. In the post titled, ‘Split Testing with a Genetic Algorithm’, the company’s CTO explained how he used Genetify to split test several lines of text, colors of buttons and various call to action arrow configurations. Over several thousand page views, the algorithms eventually determined which combination of elements resulted in the highest success rate. Read more about how the use of genetic algorithms increased their sales figures in the article.
After poking around a bit, I learned that the software has been around in some form for a while now. It was set to the back burner for some time while the creators explored a more lucrative version of the concept, SnapAds. In a November 2008 post, John Resig talks about being very excited about the rumor of the Genetify code being open sourced. A quick look in the Github history of the project shows that less than one month later, author gregdingle committed the first version. A year later, we have the most current commit, which is what I’ve been experimenting with.
The most obvious use for this code seems to be for advertisement landing pages and product sales pages. That being said, I’d love to find a use for it in my freelancing. Being a web developer first, and barely a designer second, it’s a very intriguing to think that I could allow a site’s user base to determine a very important design decision. I’d be very interested to know what web designers think of this use of genetic algorithms.


As a computer science nerd, I enjoy dabbling in all things related to web development and programming. Be it wrangling HTML/CSS across several browsers or harnessing the power of Objective-C while developing iPhone Apps, I enjoy a challenge. 


