its add van fang is that it doesn't have overhead. they're not induring those costs. so, in a sense, maybe they perform well in markets where you don't have very high utilization, but you have some high utilization that an aco could extract a pointer to. in some markets, those models will be mas, some it may be acos. there's a whole market there and for the moment, i'm going to lay that to the side. >>> now, atribute in addition to having low overhead, has this element to it. in attribution, they can wonder around. so you have to keep the patient satisfied. that's an important element of what the contribution requires. you have an na program and one question is why would you replicate that somewhere else. if you've benefitted, they're willing to actively make a decision. that might be a road that could be discussed. and the last thing i'll say, and he made this point, i think it's pretty important to keep that in mind and to make these critical, you know, make them workable models. if you had a strictly enrollment model, i'm not sure acos would be able to get enough m