This sounds like a useful way to mix in statistics and get useful approximations. I'm reading one of the survey links and it's approximately eye opening.