In class, we’ve started to study network effects and externalities.

In this post, I’m primarily looking to explore two questions. Firstly, how can we better describe network effects in terms of quantifying them? Secondly, what would it look like to filter the first question through a web lens?

It turns out that the process of quantifying the value a network effect offers is not yet completely understood. The model we’ve discussed in class follows from an intuitive derivation that layers a network effect atop a supply and demand graph; similarly, Metcalfe’s Law can be simply and intuitively stated “the value of a network is proportional to the square of the number of users of the system”. Several revisions have been made to the law, but none that change the fact that the network’s value can be considered ‘Big O’ proportional to the square of the number of users.

Much of the difficulty in quantifying the value of a network effect is based on the difficulty of defining value. For example, when trying to apply Metcalfe to the semantic web, value must be related somehow to the value of the objects described on the web and the (number of) links that connect them.

Reed’s Law plays with this ‘what-is-value’ dynamic by making a distinction between the number of possible pair connections in a network and the value of the the network itself. Reed looks at the value experienced by a user connected not to just the larger network as a whole, but also to many subgroups. Value is estimated to be higher than with Metcalfe’s Law. In contrast, Odlyzko and Tilly have developed a model suggesting Metcalfe overestimates the value of adding connections by pointing out that users do not find all (or even most) of the n connections valuable, either in the network as a whole or in most subgroups.

The following graph shows how each of the above-mentioned theories value networks with a network effect (larger version):

Moving on to the second question, it is immediately clear that network effects play a large role in the success of modern web applications. Indeed, network effects are the real secret sauce of web 2.0. Some commentary, however, suggests that in assigning value to the social, BYOC (bring your own content), web-2.0 web, the theories of Metcalfe and the others above do not hold. They suggest that such laws do not allow for the nuances or levels of interaction and features within the modern socially-enabled application. For example (from Stutzman’s The Network Effect Multiplier, or, Metcalfe’s Flaw):

I’ll try to illustrate a comparison. Indeed, Myspace’s network provides two options to you - you can either join or not join the network. If we wanted to apply Metcalfe to Myspace, this is where we’d stop. However, the value in Myspace is much more nuanced than simply being on the network; you can take value from the many things you can do on the network. The network offers a myriad of associations, including friending, grouping, messaging, browsing, stalking - actions that create a compound value that is unique for each network entrant. Indeed, each new entrant to Myspace offers others in the network the chance to create these relationships, but these many types of relationship create a value continuum - which is different than a value binary.

In a similar way to our exploration in class of a model economy without network effects, Stutzman attempts to separate what he calls the ‘core’ or ‘real’ economic value of a web service from the value of its network effect. He uses as an example the (as it has been called) massively multiplayer online photo sharing site Flickr. The core value of Flickr would be that it is a very high-quality image host and archive service. The network effect derives from the fact that Flickr is socially enabled: it offers comments, tags, favorites, messages, contacts, individualized page views, etc.: the perfect example of a BYOC architecture defining web 2.0.

Stutzman suggests that the true value of Flickr is its core value multiplied by the network effect (what he calls the network effect multiplier). He is quick to point out that Flickr is relatively light-weight in its network effect multiplier compared to services like MySpace; he suggests his thought experiment exhibits a balancing function in that MySpace’ large network effect multiplier is multiplied by a pretty small core value: a small personalized home-page.

As a brief but interesting side-note, it is intriguing to hear Eric Costello (founder of Flickr) talk about Flickr’s roots and subsequent growth. Flickr grew out of a massively multiplayer online game called The Game Neverending before becoming a real-time photo sharing and chat application built in Flash. Photo storage was added next, and quickly became more popular than the social aspect; it wasn’t until more recently that Flickr became socially-enabled once again with features like commenting, tags, and more atop its photo archival core.

In exploring network effects on the web, it is interesting to come across discussions of negative network effects. Negative network effects might result from resource limits (such as congestion on highways) or provider complacency (such as with Microsoft’s OS). In regards to negative network effects on the web, Dion Hinchcliffe writes in Hacking the Web’s Network Effect about those individuals tempted to exploit the power of network effects. The stronger the network effect multiplier (to use Stutzman’s term), the greater the frustration and damage spammers and their kind can cause us. His graph displays this nicely (larger version):

The value network effects yield is still not completely understood, and the socially-enabled web is still in its infancy. What role will network effects (positive and negative) play in the web’s near future?

Your thoughts? (Go on, get involved with the BYOC, socially-enabled, network-effecting, expert-voices, Cornell-204 thing you’re reading right now by adding a comment!)