@yaost offers some good advice in this article. In particular, there are some testing "elephants in the room" that he points out. These include the fact that the Z score used by testing tools assumes a normal distribution -- a bell curve -- in the data. That is often not the case.
He also recommends ending tests on week boundaries. If you start a test on Friday and end it on a Tuesday, you may be exaggerating your results as Tuesdays are often bigger conversion days for ecommerce sites.
This article is just nerdy enough for most marketers, and you're going to have to add some nerdy to your repertoire if you're not already there.
Group items tagged 
He also recommends ending tests on week boundaries. If you start a test on Friday and end it on a Tuesday, you may be exaggerating your results as Tuesdays are often bigger conversion days for ecommerce sites.
This article is just nerdy enough for most marketers, and you're going to have to add some nerdy to your repertoire if you're not already there.