Won't this approach have a FOUC-type experience, where the A/B div disappears after being shown to some users?

I think it's cool you're attacking this.

I think you should check out Google Web Optimizer (iirc), which will do some of the heavy lifting for you.

For example, there's no stickiness here, which I think is a typical best practice.

I love this concrete example of practical coding to do something quantitatively and empirically valuable. Yes, there are great analytics packages, but sometimes you need to do some custom analysis

Great post.

Have you played around with some of the Rack based gems out there like Split?

*https://github.com/andrew/split

Thank you! After deciding to go with Jekyll for my website I was wondering how to do A/B testing. This is perfect.