"The problem is that while KIPP is probably the fastest way to improve kids performance on Things That Can Be Measured, there's a lot of research showing how this approach is really bad for kids."

I don't understand this point here. If you cannot measure something, how can you determine what is an improvement and what is not?