While what Google is doing with their data-driven design decisions is fascinating, remember that you cannot do this for all industries. You can't give people cars in different shapes and colors and see what happens. You can't design 20 iPhone models, make 100,000 of each design and try to sell them to see which one to carry on. A good designer's intuition will tell you which 19 to nix before they hit production. And if not 19, then at least 17. My point is that Google and Amazon (and other web-based retailers/content providers) are somewhat unique in this respect.
Absolutely, but the cost is too high. You need to gasp gather people in a room, give them your product and somehow record their reactions. Don't forget that it takes real work to pick out focus groups since the dozen individuals you do use are supposed to represent your entire market. On the other hand, Google is dealing with hundreds of thousands to many millions eyeballs at once. And they have a definite metric: did the person click on the ad? Yes, using the wrong color blue is going to cost them a thousand clicks during the trial, but now they have the perfect color for the next year or two.
If you need a focus group to make up for a lack of taste, none of the 20 will be as good as 1 from Pininfarina. It's pure luck, like monkeys throwing poop on the wall to see what sticks.
There's an entire PG essay on the subject of taste and why American car companies lack it:
