That was my exact thought as soon as I read that line. At the same time though, it is important for google to realize that a/b[/c/d] testing referred to in the article will only get you to local maximums (as much as it can be said that designs catering to users can have optimal results) and to get a real leap more significant changes would have to be made.

Obviously testing major design changes with any of their core products is going to take alot more than a designer or two saying this or that design is better. I'd imagine the design process for smaller and new projects is much more open to changes on a designer's whim.