[disgruntledphd2]

Just to reiterate, this violates the assumptions under which you get your p-values.

I want an A/B testing tool that won't let you see results until it's done.