Statistical fallacies in cyberspace

The blogosphere has been set alight (don't you just love the word blogosphere?) by a report that first appeared in ZDNet about how iTunes is catching up with the "illegal" P2P networks. The report states that iTunes now has a total 1.7 million users, as much as the 2nd place service Limewire, and closing on first place WinMX. What has generated a lot of comments is the fact that the study only mentions WinMX, Limewire, Bearshare and Kazaa, while ignoring EDonkey, which boats 4.1 users at any time. The list can be found here.

The problem that I see with this report, and many other similar reports that get their facts wrong, is a combination of many different factors. Firstly, I it seems that some statistics about online trends and phenomena displays some ignorance about the target technology. A study about P2P traffic that ignores the largest network, and the emerging technology such as BitTorrent seems ultimately flawed.

Secondly, the problem with statistical analysis of the internet is the fact that it is an entirely different network, at a scale that we have not seen before. Certain survey tools, such as representative studies, will simply not show the real trends taking place online. The reason for this is that the internet is a power network, where the trends are set by a small number of websites and hubs. Market analysts will have to change their methods adequately.