jmccormac - 4:56 pm on Nov 20, 2011 (gmt 0)
Yep. The inclusion of .tk is weird. I think that the introduction of new gTLDs will have a different effect on existing TLDs than some people expect. The domain usage analysis is largely in line with a survey of about 2M .co.uk domains I did a few years ago. I did a survey of 2.3M domains in October and the results are absolutely terrifying. Only 13.53% are Active/unclassified sites. Applying a more aggressive algorithm to analyse possible other TLD sites being served as .eu (no 301/302 and with other TLD hrefs) drops the Active/unclassifed percentage to 5.71%. But that's a seriously aggressive algorithm and I am still refining it. I will probably publish the results next week.
Nominet apparently sampled some sites for its usage survey rather than running a full TLD survey. This is probably ok for general indications. At least it is not as wildly inaccurate as the Afilias/INFO usage report or as optimistic as the Eurid Insights report. Sampling rather than running a full TLD survey is not a reliable way of assessing usage in a TLD because people break HTML, companies have their own versions of Holding Pages, PPC is often stealthed as seemingly genuine content and more importantly, other TLD sites are often served without 301/302s.