Website testing using Genetic Algorithms

Introduction 

Website testing, in terms of maximizing site usability, or marketability has really not evolved very much from the traditional A/B approach. Given the current state of technology we can certainly utilize more powerful platforms and applications that deliver automated solutions and/or analytics to help us (designers/web authors/front end developers) better deliver usability, branding, profits, etc through design/styling changes.

Background 

Currently, the most widespread testing approach used by companies is A/B testing. In website testing, A/B allows for two distinct designs to be presented to visitors of the site being tested. The choice of which design to present is generally random. With this type of testing, companies can benchmark which design is favored of the two, and proceed with making the design change permanent for the time being.

Multivariate testing is another approach that goes beyond the limitation of two variants. With it companies can test multiple designs/changes at the same time, and watch which one wins out. This is obviously an improvement, but again is still limited to a designer making a finite number of designs to test.

This is where GA's (Genetic Algorithms) really shine! Please read up on Genetic Algorithms to fully grasp the intent of this article. Utilizing GA's, designers/developers would no longer be constrained to single all-or-nothing designs. Implemented properly, GA's will allow a designer to control the variations of any style or design choice on a site, while still allowing for huge variation, especially if much of the design is left to the system to generate. Ultimately, an optimal, or near-optimal design will be produced after a period of testing with this approach. Furthermore, as style, fashion, tastes, etc change, so too can the genetic algorithm to adapt to these new variables in the enivronment. 

Benefits of Windows Azure 

This project will be service based, allowing designers, programmers, etc to subscribe their website, and its design (stylesheets/html) to the service. In order to fully automate a system of this nature to allow for a vast number of sites and bandwidth, we need to employ scalable cloud computing. Windows Azure is perfect for this task; using Windows Azure the service will be able to cope with growing needs by creating additional resources on-demand.