It sure looks like, at an increasing rate, kids are choosing KA. After all they are held most directly accountable for their academic success or failure. Wouldn't demand then be a reasonable gauge of value?
Students might say that method A "feels better" or "is more fun" than method B. However, it is entirely possible that method A is worse because, in the end, students will actually not have learned what they would learn with method B.
This is not to say that students' opinions are irrelevant. After all, when students don't enjoy what they're doing, they will often block out the lesson and not learn anything due to the lack of fun.
Still, it is pretty absurd to think that students could ever be objective in evaluating the teaching methods they are subjected to.
It's easy to recognize good teachers. They are first, understandable and second, motivating. If a teacher is not readily understood, or if they do not motivate the student to learn, then they are not very good. Children as well adults naturally use these metrics to evaluate not just teachers, but communicators in general.
Fractions and decimals are not difficult. I doubt that kids who have been learning math with Khan have problems understanding them. But ìf Khan has been a little unclear for some percentage of students, the addition of a few exercises should solve the problem. The lacunae of a smart teacher are easily fixed, unlike the confusion caused by a teacher who doesn't understand the material.
You must realize that this is an incredibly simplistic view of learning? How students learn is an extremely complicated process. While decimals are not difficult, understanding exactly where students struggle with the concept and determining a sequence of learning activities to help them overcome that concept is extremely difficult.