It was quite simple. We had 6 tabs at Intuit.com. We had an A/B/C test. Control was the nomral order, Tabs in one alternate order, Tabs in another order. (No other changes) The last tabs had an abnomral bump in click through rates against the control. The bump wasn't super large, under 5% higher; still it refuted the idea that the second tab would get the bump which is what the common wisdom is.