Does ping data show power law applies to weblog market share?

FRANCE: Jean-Luc of mediaTIC referred me Stephane Le Solliec's study of blogfarm share using weblogs.com data.

From bsentinel, pings by source:

1. 46390 (12.75%) : blogspot.com
2. 37921 (10.42%) : modblog.com
3. 32563 (8.95%) : persianblog.com
4. 23255 (6.39%) : blogdrive.com
5. 11604 (3.19%) : livejournal.com
6. 6651 (1.83%) : 20six.fr
7. 5565 (1.53%) : myblog.de
8. 4776 (1.31%) : co.uk
9. 4558 (1.25%) : tblog.com
10. 3603 (0.99%) : csdn.net

These from 363,889 weblogs.

Please pardon my poor translation from French.

28 June 2004

For the last two or three weeks I looked at all the blogs noted as updated on http://www.weblogs.com/ and I compiled them by hand (finally, almost) in a small list.

The small list makes now more than 130,000 blogs.

To see which were the most used farms with blogs, I created a small page which lists, by fields, the blogs of my list: shares of the blogosphère.

One can notice:

• blogspot is first with almost 15% of shares to go. Ok, not bad.
• persianblog in second. Large surprise. Considering its alphabet, impossible for me to really know. Are the Arab language weblogs in another country, although blogs of Iran?
• 3rd, 4th and 5th: blogdrive, livejournal and modblog (I did not know this last)
• 6th: 20six.fr

20six.fr largely replaces joueb.com, the first French-speaking platform of blogging, three years' old. U-blog.net is difficult to measure, because only the paying blogs can automatically ping www.weblogs.com (the free ones must to do it manually).

What is also interesting to see, is that of all the 20six platforms, the French site is in first place. The German site has a third of the blogs. Cock-a-doodle-doo! France (or rather French-speaking people) seems to be completely many bloggeurs.

As regards TypePad, which is the most advanced solution that I know, it finds 1 percent, which is not so bad since it is also the most expensive solution, in a world where the free ones are almost omnipresent.

At the end, these statistics are to be taken for this what they are. As I explained at the beginning of this post, I only entered blogs which ping http://www.weblogs.com/. This excludes all the blogs that:

• do not have this functionality
• didn't want to activate this functionality (common in private blogs)

All the same, it is a reliable indicator and in real time.

Conclusion: The blogosphere is incredibly dispersed. The first 100 sites represent only 37,000 blogs [28%] out of the 130,000 of my list.

P.S.: The next stage will be to scan all the blogs to extract the GENERATOR tag to know which software are most used.

P.P.S.: A small coat of paint on bsentinel would be good, but I think other tricks are to be made now.

Jean Luc writes:

It is far from obvious how to evaluate blog system share.

Elise Bauer produced an analysis of the shares of market of blogs in this post: "An Overview of the Weblog Tools Market" while basing itself on this methodology:

"To try to get an answer to this question I've turned to Google. By typing in the domain name of a tool you can find the number of web pages that link to the domain name and the number of pages that contain the search term of that domain name."

The results are surprising in more than one way. As it is indicated in the comments, one notes the absence of B2 evolution, a blog system well-known and used (more especially as B2 is not new) but also of very dynamic platforms such PersianBlog, Polish solutions or the whole of the blogs 20Six (France, Germany, Great Britain, Netherlands) which carry a considerable weight.

Like Phil Wolff specifies in an article for BlogCount, it is significant to wonder about why certain solutions of blogs are better referenced in Google compared to others.

This analysis of Elise Bauer is thus partial (as she specifies it) but as one of the competitive metrics the blogging market is to quantify, qualify and to characterize these market shares, it is included/understood well that the stake is serious.

It would undoubtedly be necessary initially and easily to cross these figures with other search engines... Cross analyze also with BSentinel de Stéphane Solliec which screens more than 350,000 blogs in the world. JLR

Unlike Google, bsentinel reflects active weblogs.

My bsentinel wishlist:

1. More ping servers. Many ping servers are used today, at least a dozen widely. Include more of them, eliminating duplicate pings.
2. Collect time data. When do Xanga users post vs. LiveJournal users? How are shares changing over time?
3. Collect post-level data. Weblogs are one thing, but posting activity is another. What is the market share by post? How often to Diarylanders post vs. Blogspotters? TypePad.fr vs. TypePad.de?
4. Check for sub-blog pings. I use Radio UserLand for one of my weblogs. Each post belongs to one or more categories, seen on my web server as a separate weblog. Radio pings for each updated "category" even though I only published one article. The clue: 3 pings in 10 seconds from the same weblog.
5. Check for "update" pings. If a second ping comes from the same blog in 3 minutes, it is probably just typing and not a new post. What percent of pings are updates?
6. Spider. Other forms of discovery, like crawling, are more expensive (time, bandwidth, computation) but yield more data. Generator tags, as you said. Sometimes language and geography data.

My guts tell me there are around 10 million active weblogs in the world. So bsentinel is picking up about 1 percent. Does using ping data introduce much bias?