Comments on: Yahoo Mobile Search Results Blow Google Out Of the Water; But Why Can’t Both Better Their Service? http://www.msearchgroove.com/2008/02/01/yahoo-mobile-search-results-blow-google-out-of-the-water-but-why-cant-both-better-their-service/ At the Intersection of Content & Context Wed, 17 Mar 2010 18:31:15 -0400 http://wordpress.org/?v=2.8.4 hourly 1 By: Steve Ives http://www.msearchgroove.com/2008/02/01/yahoo-mobile-search-results-blow-google-out-of-the-water-but-why-cant-both-better-their-service/comment-page-1/#comment-2589 Steve Ives Thu, 21 Feb 2008 09:06:28 +0000 http://www.msearchgroove.com/2008/02/01/yahoo-mobile-search-results-blow-google-out-of-the-water-but-why-cant-both-better-their-service/#comment-2589 We've also run side-by-side comparisons of Google Mobile vs Yahoo OneSearch (also Microsoft, and also Google on the iPhone vs Google on the N95 - but that's another story). Our findings for a similar number of search terms are that Google and Yahoo are very similar in relevancy performance, with Google slightly ahead, and both perform significantly better today than MS. My feeling is that the differences in our results can be at least partially explained by statistical effects. 200 searches on 2 phones is nowhere near big enough a sample size to be statistically significant. You need at least a sample size of at least 2,000 before you can begin to say that the variation in the experimental results is statistically significant. We’ve also run side-by-side comparisons of Google Mobile vs Yahoo OneSearch (also Microsoft, and also Google on the iPhone vs Google on the N95 – but that’s another story).

Our findings for a similar number of search terms are that Google and Yahoo are very similar in relevancy performance, with Google slightly ahead, and both perform significantly better today than MS.

My feeling is that the differences in our results can be at least partially explained by statistical effects. 200 searches on 2 phones is nowhere near big enough a sample size to be statistically significant. You need at least a sample size of at least 2,000 before you can begin to say that the variation in the experimental results is statistically significant.

]]>