Hacker News Reader

I wonder whether they split test these types of things?

They probably do, but Google is unlikely to be immune to corporate pathologies, where for example it is politically impossible for some team to leave well enough alone. I suspect that could easily end up trumping split test results.

View on HN

I believe they do a tonne of testing, this video from Google is a great illustration of their process:

http://www.youtube.com/watch?v=J5RZOU6vK4Q

Isn't Google incredibly famous for their dedication to A/B testing?

A/B/N testing from what I heard. One of the examples I know of it that they split the users into about ten groups, and very slightly changed the blue hyperlink colour between them. They found that even a slight variation made a difference to the click-through rate.

I've had this same experience at work. I've complained about how ugly some features of our sites are on several occasions, only to learn that they are that way because the ugly (version/size/whatever) performed so much better than the others in our tests.

Other times the differences are subtle but increase the click-through rate by surprising amounts.

Reply on news.ycombinator.com