Tech stock followers, not to mention bloggers with a taste for schadenfreude, were entertained last week by reports of a research paper out of Princeton predicting aEUR" well, sort of predicting aEUR" the decline and fall of Facebook. The paper, by two grad students in the engineering department, probably doesnaEUR(TM)t mean a whole lot; simulation models are no better than their assumptions, as FacebookaEUR(TM)s in-house statisticians were eager to point out. But itaEUR(TM)s worth a bit closer look, if only to remember how unstable markets built around information technology and dependent on the magic of networking really are.
The idea is clever and straightforward: In many ways, the process of social networking resembles that of infectious disease transmission, and thus can be modelled in similar fashion. The ebb and flow of communicable diseases turn on a variety of factors including population mobility, the latency period before symptoms appear, the method of communication (air, water, etc.) and ease of communicability aEUR" yadda, yadda, yadda. The path may prove highly predictable with modest amounts of data, or almost unpredictable with lots of (think butterflies and hurricanes). But with the epidemics we worry about, the infected population generally grows until it is interrupted by intervention (quarantine, vaccination) or by its own internal dynamic and then collapses.
The Princeton guys applied the very simplest SIR epidemic model that puts everybody in one of three compartments (Susceptible, Infected and Recovered) and predicts how the size of the compartment changes with time using three differential equations. Lo and behold, the rise and fall of MySpace, once FacebookaEUR(TM)s archrival, followed the epidemic pattern very well. The aEURoeinfectedaEUR? population (users) gradually recover (lose interest) and as the network thins, the process snowballs.
Facebook attacked the model as aEURoegarbage in, garbage out.aEUR? A three equation, three parameter model is hardly adequate to the task. Besides, the analogy of gaining immunity to losing interest is strained. Everybody gets over a disease or dies; social networks need not function that way.
WhoaEUR(TM)s right? Certainly, the grad students aEURoeprovedaEUR? nothing. But thereaEUR(TM)s no easy way of getting around the notion that social networks are unstable, and that snowball effects of one sort or another are always lurking.
But I have my own hobby horse to ride here. IT companies like Facebook (or Microsoft), especially the ones whose products depend on network effects (Microsoft Word becomes more valuable to me as the number of users grows), arenaEUR(TM)t like old-fashioned monopolists. Their true competition isnaEUR(TM)t from similar products, but from innovations that break the network link and send them spiraling the other direction. Students of antitrust, take note.