[raverbashing]

You can try, but it's tough

Seeing Google try different shades of blue automatically, just goes to show that they don't know what they are doing

"To build a great UX, one has to step back and think about what’s most important, try to come up with the most simple and shortest path to get there, build, analyze what works and why(not), measure, test, rinse, and repeat."

This is very important. But get this: measuring and refining won't fix big errors.

So work with a designer and don't BS him trying to do "live genetic programming" your site with your A/B tool.

[badclient]

Seeing Google try different shades of blue automatically, just goes to show that they don't know what they are doing

Why does that show that "they don't know what they are doing"?

[raverbashing]

Forgetting for a moment all the theory of color and contrast /equilibrium in a design:

1 - different people see different shades (difference in monitors)

2 - Unless there is a glaring problem with the shade of color chosen (like something most designers wouldn't approve), changing shades is one of the least important areas of improvement possible

3 - The current redesigns of GMail / G+ / GReader (among others) are very showing that Google is thinking like "engineers" and not like UX experts

[badclient]

(1) Can't that be accounted for in some manner, especially when you have so much data? Typically when I design tests at work, it is not unusual to add many rules and exceptions to make sure the data is as accurate as possible.

(2) This is where it becomes tricky. You have to understand that these decisions are made at a very local level by small teams with specific goals(for example, "bump revenue from sidebar links by x%"). Also, it is unrealistic to suggest that a company the size of google can only work on super big items when at Google's volume, small improvements can add up and result in hundreds of millions in additional revenue.

(3) Not a fan of the new gmail but using their blue shade experiment as supporting evidence for why their design sucks isn't very powerful.

[raverbashing]

1 No, you really can't account that. You can't account for people's eyes as well. You may make a guess (maybe on display resolution) which is not precise at all.

2 Google does an can certainly work with whatever they want, but they should try the big improvements first. Their work with page loading times is great, and that has been shown to correlate with earnings. But trying to tune the shades when there are more important things to be worked is pointless

3 It goes to show they don't understand design. In some aspects they do (like GMaps) but it seems to be declining

[j_s]

Wouldn't #1 be an argument for testing to pick the best color (or at least a big part of any decision) - it doesn't matter what the designer picks, it matters what people see!