Metcalfe’s Law: Two Plus Two Equals Sixteen, and Sometimes More; For Unusually Large Values of Two
Novell grew fat and happy, but NetWare was egregiously poorly designed, and installing, changing and maintaining it required an expensive, trained specialist. What’s more, the network behaved rudely and inappropriately, frustrating users. Novell failed to realize this, probably because millions bought NetWare, but its customer base was motivated by need, not desire.
In the early ’90s, Microsoft, 3Com, and even Apple began to ship LAN products that were as capable as NetWare, but that didn’t force customers to depend so heavily on third–party experts for installation and — especially — maintenance. Novell watched in mute horror while its leadership position evaporated.(p75)
—The Inmates Are Running The Asylum; Alan Cooper
One of the central tenets of the new economy is Metcalfe’s law. As commonly articulated, the value of the network increases at the square of the nodes or users, depending on the source. Metcalfe observed that early tests of networks had difficulty justifying their existence. That is, until and unless some basic minimum utility could be demonstrated by connecting enough people to justify using the network. George Gilder is responsible for the wording used today.
If there is one thing I’ve learned about the new economy, it is the speed at which nebulous words like value are seized upon for mischief. I can well understand the value component of laws like Moore’s and Metcalfe’s to the bottom line of firms like Intel and 3Com, and various new economy gurus. I have little trouble understanding at some point the number of nodes justifies building the network on the basis of bare minimum utility. What I can’t understand are the more personal value metrics.
A manager for a famous national chain of computer training centers confided to me his president threatened to rip out their network every so often. The president apparently concluded the betting pools, inspirational quotes, and less savory material communicated didn’t necessarily constitute ‘knowledge work.’ For any business to have this problem is disturbing. For a business built upon training knowledge workers to compete in a networked economy to have this problem is tragic. What this says about networks of all types is debatable, but no debate is forthcoming.
The obvious response is simply to state that the value creation effect is much greater than any marginal abuses. However, the amount of abuse can be attributed in part to network structure and interaction design. Which is to say that it’s all well and good to assume network use statistically washes out, unless yours is on the bottom of the statistical heap.
The Network May Have Value But Is It Desirable?
When software implements social interfaces while disregarding cultural anthropology, it’s creepy and awkward and doesn’t really work.
— It’s Not Just Usability By Joel SpolskyEven today there are the occasional articles extolling the virtues of the company intranet. One recent example consists of how someone posted some sales tips, causing a flurry of equally good tips to be added from far flung points of the company. Given the borderline excitement the article’s author seemed to express, you’d never have known there ever was a DTP revolution. An era chock full of company newsletters with tips from the field, decades before. While the cost savings of the ’net over paper are obvious, the author should have noted the quality of ideas adopted and resulting sales increase. That this is so rarely done is far more informative of the current state of affairs than the article’s content.
Joel Spolsky calls into question the design of social networking interface designs, citing Danah Boyd’s article Autistic Social Software. “Consider, for a moment, the recent surge of interest in articulated social networks such as Friendster, Tribe, LinkedIn, Orkut and the like. … While this approach certainly has its merits because it is computationally possible, I’m terrified when people think that this models social life. It’s so simplistic that people are forced to engage as though they have autism, as though they must interact procedurally. … Furthermore, what are the implications of having technology prescribe mechanistic engagement? Do we really want a social life that encourages autistic interactions?”
When I read an article that the killer app of 3G is voice, I wonder what the killer app was from previous generations. Were we to take the same 15,000 word conversation between 5 individuals, and replicate it as text, then voice, then video, I wonder what the difference in actual realized value is. And when I see a PCS license auction and the bankruptcies it spawns from carefully calculated bids, Metcalfe’s law an element in most, I have to wonder.
Clearly, value doesn’t just happen. It is what you do with a connection to the network that counts. It seems to me that Metcalfe’s law, while useful, is fraught with the potential for folly. There seems to be a tendency to build networks, and equip them with technically clever features, then wait for value to show up automatically. While desirable interaction design and service policies don’t get mentioned in many new economy texts, the frivolous factors like actually being able to use the connection drive node count.
Resources
- With emotional accesibility comes actual accesibility — until desirability design becomes mainstream, network designs are stuck at the level of accessibility for the autistic.
- Andrew Odlyzko and Benjamin Tilly, University of Minnesota researchers, find Metalfe’s law “is a significant overestimate”. Their network effect law, in contrast to Metcalfe’s, shows that incumbent powers have a reason to shut out smaller new arrivals.
- An interesting Cringely article on the problems of bare minimum access, a problem far below the threshold implied by Metcalfe’s law. Also interesting is Xerox 1974: “[Ethernet] would be a failure” not least for the historical perspective, comparisons between networks and the difficulties people have when comparing the advantages and disadvantages of networks.
- Social network analysis maps links between people on a network. Troubling is that mapping links rarely tells you when the connections and flows are antisocial and dysfunctional. Add what Martin Ruef, assistant professor of strategic management at Stanford has found, which is Innovators Navigate Around Cliques. “Contrary to common assumptions,” says Ruef, “the evidence suggests that in many cases strong social ties do not provide significant new information, so it helps not to be as embedded in them.” Put these pieces together and begin to discuss what information based social networks should look like.
- Emergent Music’s rating model has a methodology for creating value, reducing misinformation and disinformation, and points for the ability to predict which can identify the opinion leaders important to social networks (or coolhunters important to marketing). Movielens is a similar project developing its own analytics.
- Get smart, work a little less wired makes the argument more connectivity distracts from productive work. If not all connectivity is good connectivity, the case for value simply appearing by adding nodes or users gets harder to make.
- Robert X. Cringely. Mentions the 2G bankruptcies and 3G licensing computations ($576 Billion) using Metcalfe’s law. Also explains the advantage of 3G. Noteworthy with Nextwave’s $4.7 billion dollar license bid woes making news now. This note could be updated every few days with new telecom failures. The reason given for this is ‘overbuilding’. The more informative reason would be the kernel which formed the conceptual basis for the overbuilding. (Hint: The recession was an accelerant, not a cause.)
- Fortune Magazine Who wrecked telecom? Critics blame the fledglings. But the folks who did the most damage were veterans.
- Concurrency means the free lunch provided by Moore’s law is over.
- When given a choice between chip densities doubling or total lower prices, something tells me Walmart Trumps Moore's Law.
- Occam’s Razor isn’t looking too healthy, either Geoffrey I. Webb, a computer scientist at Australia's Deakin University, found problems in experimenting with software employing both assumption of similarity and Occam’s razor. His findings have major implications for the assumptions governing data mining applications. Occam’s Razor: the simplest solution is the correct one, all things being equal. I must confess, I favor Einstein: things should be made as simple as possible, no simpler.