gmoney - 7:48 pm on May 31, 2003 (gmt 0)

Hi Julia,

Just to clarify my thoughts above, I believe the two main sources of errors with using Wordtracker (and WebCEO) data to estimate things on other search engines are small sample size (i.e. the database is only a small fraction of the web) and different demographics (i.e. maybe MetaCrawler searchers have different interests and search differently than Google searchers). Anyway, when I wrote my post above I was primarily thinking about errors arising from different demographics.

I agree that the thinking in my earlier post did not apply to your example because I believe the problem you mentioned in the post above has more to do with the small sample size magnifying strange behavior than it does with the potentially different demographics of Meta search engines vs. others. I believe you would see similar problems even if you used a small sample size of Google searches (i.e. say you could track 1% of all Google searches for a day) to estimate Google traffic for a month.

It would be interesting to see the details of how WebCEO tries to filter strange anomalies out.