Wednesday, June 18, 2008

Yahoo! getting screwed by a regex bug?

[Update: Yes, Compete says it was a bug.]
I hope not!

SearchEngineLand says Compete uses the following regex to count the number of Yahoo! searches:
^/search\?(.*&)?p=[^&]+

It actually excludes a vast number of searches that uses Yahoo! link tracking parameter.
For example, a search that I made from my Firefox searchbox has the following url:

http://search.yahoo.com/search?p=Perl&ei=UTF-8&fr=moz2

No issue here. But, when I refine the query by adding "language" in the search box in the results page, following is the url:

http://search.yahoo.com/search;_ylt=A0geu6Xv_FhIOyoBokNXNyoA
?p=Perl+language&y=Search&fr=moz2&ei=UTF-8


Anybody who has worked with regexes can see that this url and scores of similar urls of query refinement won't match the regex supposedly used by Compete and hence not counted! And, all this while, Yahoo! stock keeps taking a beat for failing in Search. If true, it could be a multimillion dollar bug.

Anybody from Yahoo!, Compete who can confirm or deny this?

Meta: If you're posting this, link back to this post, please.

Labels: , , , , ,

2 Comments:

Anonymous Anonymous said...

Hi,

Steven Willis here from Compete. I can verify that these are some of our search engine query rules, and as a result, we might be missing some yahoo search queries. I can also tell you that in other areas of our metrics we actually take this semicolon URL from yahoo into account. So, for example, if you were to go to the yahoo results page with the semicolon and and clicked on the first search result to perl.com we would definitely capture it as a search referral from the term [Perl language] to perl.com.

Thanks for the heads up. I don't know how much of a difference it'll make in the search engine market share, but we will be adjusting our search query rules and re-running our numbers to see what the effect is. Keep an eye on our blog.

-Steve

P.S. I sure do wish the market would make million dollar swings on our word alone :-).

June 19, 2008 12:59 am  
Blogger Sundar said...

Thanks for the comment, Steve. I'll be waiting to see the results from re-running the numbers.

- Sundar
P.S. I liked the pun in your postscript. :-)

June 19, 2008 8:30 am  

Post a Comment

<< Home