Early in December I wrote about Content Experiments by Google. It’s a great way to maximize conversion rates of your web pages, which increases profit. It compares up to five different versions of a web page and calculates which one works the best. Well, Google has a variation of this Content Experiments. Google calls it the “multi-armed bandit experiment”.
This variation has the same goal (maximizing profits) but works a bit differently. Instead of sending customers randomly to your different web pages, this actually has some math behind it. Don’t worry, I’m not going to go into the technical side of it but if you would like to know you can go here (under statistical validity). Essentially, this experiment will start sending more customers to the more successful web pages, or the web pages that have higher conversion rates than your other ones, narrowing down the pages served to the ones that produce the best results.
So if you have five web pages you are testing and one of them are just not working out for you at all, you may get less than 5% of all your customers directed to that page. If it’s producing terrible numbers, the experiment may just never send traffic to anymore. So, why would you use variation of Content Experiments over the other? Well, you will get results faster for starters because the poor performing web pages will get less and less traffic directed to them. Since this is keeping your traffic off of the less successful pages, it is more cost effective since it is sending customers to the successful variations more often. Also, you will be able differentiate the highest performing pages from the other pages.