August 18, 2006

The Value of Metcalfe's Law

For whatever reason Metcalfe's Law has been all popping up everywhere I look over the past 24 hours.

The July issue of IEEE Spectrum has an article "Metcalfe's Law is Wrong" that is probably the crystal under which all this attention can be formed.

Metcalfe's law basically states that a value of a network grows exponentially with the number of nodes. A telephone network with only one phone is worthless. One with two is usefully only to the few people who can reach those two phones. But a phone network with a million phones has a massive value. The authors of the IEEE article don't dispute the existence of "network effects", the fact that as nodes increase networks rapidly increase in potential value. What they dispute is the exponential math that Metcalfe, the inventor of Ethernet and founder of 3Com, uses. The authors instead argue for logarithmic growth, or something close to it. They are probably right.

They also make an important point that "Metcalfe's law" is not a law at all, but merely a empirical observation that serves as a useful guide for analysis. What they completely miss is the wide open flaw in Metcalfe's Law, which can be summed up in one word, value.

What the fuck does it mean when you say the value of a network increases, with the number of nodes? Well it depends on what sort of value you are talking about. A network of phones in a world of deaf people has no value no matter how many nodes it has. What a network gains when it adds nodes is not value but potential value. That potential value only can turn into "real" value when valuable content flows through the network from node to node. You just can't sit down with Metcalfe's law, in either the original exponential or new logarithmic version, and use it to calculate the value of your network based on the number of nodes. All those nodes are worthless if they don't have anything to say to each other.

You can build a network, you can get a guideline of it's potential value, but to unlock it you first need to figure out just what value you are talking about. And then you need to get that value into the network in a way that is communicable and then and only then will the network actually begin to produce the values you want.

update: not 10 minutes after I posted this I stumbled on Metcalfe himself responding to the IEEE article in quite an interesting manner. And as usual Fred Stutzman has some interesting things to say as well.

Posted by Abe at August 18, 2006 11:29 AM

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)