But, I'm going to call foul on Ross' handwaving citation of Reed's Law as an antidote to Sarnoff's broadcasting formulations. Let's get one thing straight: Reed's Law is obviously, trivially, and blatantly false. It posits that the utility of a (human) network always grows as its membership grows. In fact, the utility is not just O(N^2), but is instead O(2^N). Umm, hello? If there aren't any negative terms here, then all groups naturally grow without bounds. That obviously flies in the face of both common experience and common sense. There is a conserved item in this game - time and attention. That's the basis of competition between old and new media, and among new media forms. A 'Law' that can't even model that fact is a bit of a waste, IMO. Try again.
I'm not one to quibble with Tim's not-so-humble opinion, and this is the kind of post that wakes the trolls, but consider this -- each connection isn't solid state, its an option within a network that changes over time. Under Sarnoff's Law, the value of the network is the number of subscribers. Even subscribers have scarce time and attention. When I subscribe to your feed its no guarantee that I will actually read your mind-bending drivel. But I always have the option to consume what you produce.
To me, Reed's Law is a measure of the connections between groups with each connection being a real option. People exercize options when its worth paying the price, sometimes in time and attention. People who span or bridge groups are extremely valuable in this regard. They serve as a proxy for groups to save them time and attention, but often out of the self-interest of information arbitrage benefits. If the arbitrage spread grows, other members of the group will invoke their options for more direct connection. Now groups come and go, but the latent potential of some of those connections remain. Kerry and Spirit of America stands to benefit from ties created by the Dean Campaign.
Now I'm not suggesting that the model to replace CPM is a combination of Black Scholes and Reed's Law. Too complex, and they are models, not measures. What's different with new media is simply that its not the number of impressions you make, but who you impress.
In other words, instead of subscription counts, its the number of subscribers my subscribers have, discounted by the probability of my memes getting through. Cost Per Influence.