Nicely done. One problem I've found with optimizing for revenue per visitor is it takes ages to get significant data unless you're right at the end of the conversion process such as in your example. Even then, it's difficult to justify testing on a slice of the users (by campaign, for example) because it may take a month or so before any meaningful answers come in.

Are you planning on integrating additional measurements later on? What if my treatment produces more initial revenue but the lifetime value of those customers is lower by a negating amount through additional support, returns, chargebacks, and reduced rate of future re-orders, etc...