i have read 2 interesting studies in the last couple of days. the first is a release from Comscore which details people’s clicking habits on display advertising. it seems that heavy clickers (6% of the population) account for 50% of all clicks on display advertising. i personally don’t click on banners, but there must be someone who does so (and apparently this person is below average income, a heavy internet user and into gambling, auction and career guidance sites). however if you are basing your roi on clicks alone and not any post click behaviour, you’re in deep trouble anyway.
the other study was a definitive guide to online testing from some ex-amazon/microsoft guys. they talk through the relative advantages and disadvantages of the various approaches; A/B testing, multivariable (not multivariANT), parallel and split tests. its well written and has all the statistical backup of why you shouldn’t only rely on the HiPPO (highest paid person’s opinion)