I agree sorta, but only for the more tweaky visual design aspects, rather than interaction design.
If anything, I see Google as actually less data-driven in its design than many areas. Industrial psychology and HCI are traditionally very big on data-driven design for interaction design, with a big focus on Methodologies, quantitative measurement of task performance improvement, and various other statistical evaluation metrics. But the main downside of that is that you can't really answer open-ended design questions: you can A/B test between two specific proposed interfaces, and determine A causes N% of users X quantity less hassle than B does (or whatever your metric is), but if you want to invent an entire new interface, you're basically doing a random walk in design space, which will take a really long time to get anywhere. I mean, imagine how many A/B questions you'd have to ask to invent the Gmail interface, using, say, the Hotmail interface as your starting point, and having to justify every change you make from that starting point using data.
I suppose it depends on your norm, though--- Google is very data-driven by the web-design standard, but sort of loosey-goosey, anything-goes compared to the much more statistics-and-methodology-heavy design practices that go on in traditional engineering and HCI.
If anything, I see Google as actually less data-driven in its design than many areas. Industrial psychology and HCI are traditionally very big on data-driven design for interaction design, with a big focus on Methodologies, quantitative measurement of task performance improvement, and various other statistical evaluation metrics. But the main downside of that is that you can't really answer open-ended design questions: you can A/B test between two specific proposed interfaces, and determine A causes N% of users X quantity less hassle than B does (or whatever your metric is), but if you want to invent an entire new interface, you're basically doing a random walk in design space, which will take a really long time to get anywhere. I mean, imagine how many A/B questions you'd have to ask to invent the Gmail interface, using, say, the Hotmail interface as your starting point, and having to justify every change you make from that starting point using data.
I suppose it depends on your norm, though--- Google is very data-driven by the web-design standard, but sort of loosey-goosey, anything-goes compared to the much more statistics-and-methodology-heavy design practices that go on in traditional engineering and HCI.