Long Tail Search Engine - Google Search with a Tail

This blog was set up to write about and seek feedback for The Long Tail Search [a.k.a SearchTheWeb2, TYNTO! That You Never Thought of!] Make a Comment or Suggestion? ...Thank you for stopping by.

Blogger

Blogger

Blog Archive

About Me

My Photo
View my complete profile

Sunday, August 5, 2007

Zipf's Law for the Global Search Market Share & Rankings

This post is intended to propose a new method to estimate the Global search market share for all search engines based on the Zipf's law. The estimated share for the US search market has been provided and updated monthly by several sources. For example, the estimated share of the number of searches that happened in the United States in May 2007 as shown in the below table.

SEngines
Google
Yahoo
Microsoft
Ask
Others

comScore
49.50%
25.10%
13.20%
5.00%
7.20%

NetRatings
56.30%
21.50%
8.40%
2.00%
11.80%

Hitwise
65.13%
20.89%
8.40%
3.92%
1.66%

Average
56.98%
22.50%
10.00%
3.64%
6.89%


The above table also gives the average share for each major search engines based on three sources. Based on these average share percentages, the search share for other search engines (other than the big four) are about 6.89% for the month of May 2007.

For the Global Market, the statistics are not this simple. Quite a few numbers have been bouncing around lately in terms of search share. On a global basis, comScore estimates Google held a 65.7 percent share of the Web search market while Richard Zwicky at Enquisite reported that Global share was about 73.8 % with excluded US and Canada markets.

Zipf's Law Application
Zipf's law can be presented by a plot with the axes being log(rank order) in the x-axis and log(frequency) in the y-axis. If the points are close to a single straight line, the distribution follows Zipf's law.

If we assume that global market share for all known and unknown search engines in the world is followed Zip's law, what should this global market share be?

A graph of 4 major (Google, Yahoo, MSN & Ask) and 100 alternative search engines selected and maintained by Charles Knight is shown as below.



The graph was constructed by plotting log(market share) vs. log(rank order). Market share for each engine is based on various sources of estimations and then normalized based on the Google share that can be assumed at any percentage. A constraint will be that total of all search shares from these search engines should be at 100%.

The result graph shows that the data points are not on a straight line. The reason of this behavior is that there are many missing search engines in the data set. Although most of these missing search engines can't be known but some such as Baidu, AOL, Yandex, AltaVista, Naver, etc. can be added to the data. If these search engines are added, a straight line with a slope of -2 is clearly taken place. As Charles Knight mentioned in his blog, the top 100 alternative search engines have been selected among a data base of 1000 engines. Fortunately, not every search engine needs to be identified in this study. A plot can be constructed without their identities as shown below.

Conclusions
Zipf's law can be used to approximate the Global search market share for all known and unknown search engines. Global search market share currently follows a Zipfian distribution with slope of -2 and an intercept point (Google share) at 60.5 % as June 2007. There are approximated 1000 search engines in the world with a daily search traffic of 880 millions.

Posted by Long Tail Search at 9:15 AM  

Labels: Alternative Search Engines, Global search market share, Search Ranking, Zipf's law, Zipfian distribution

0 comments:

Post a Comment

Newer Post Older Post Home
Subscribe to: Post Comments (Atom)

Featured Posts

  • Global Search Market Share: Zipf's Law New Research
  • Ranking: Top 100 Alt Search Engines New Update
  • Links

  • Google News
  • Long Tail Search
  • ________________

    SearchTheWeb2 Community

    ________________

    Long Tail Search Engine

    ________________

    Add to Technorati Favorites

    ________________

    ________________

    Clicky

    ________________ /*script src=~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ */