check out this very interesting story on compete blog about web search fulfillment (ie., leading a searcher into a website). notably, searches on yahoo result in the user clicking on one of the results 75% of the time, where as the same happens 64% on google. though a yahoo-centric conclusion is that “Yahoo! more effective at getting consumers the results they want”, but it does have other interesting angles :
results shown: this data probably measures the outgoing clicks from everything on the search results page. Yahoo! includes what’s called ‘inside yahoo’ (pointing to properties owned by Yahoo) and the ‘wow modules‘ (little snippets of info from movies, music, maps, etc). these inherently will have high click through rates, as do most name-brand sites. Google tested a similar feature, but canned it due to backlash from religious search purists.
longer long-tail: google gets twice (or thrice, depending on whose number you believe) as much traffic, and their index is probably larger due to their near real-time indexing., causing lower recall and results that are probably relevant but from unknown sources. users probably (this is my guess) don’t click on links that point to websites that aren’t well-known.
search experience: yahoo’s latest search experience enhancements will probably tip the statistics ever higher towards Yahoo. if compete is measuring the page views as proxies for distinct queries, then ajax-ified results page would count fewer page loads causing the percentage referrals statistic to go up in the future.
advertising: if you are an advertiser, why wouldn’t you prefer to advertise on a site that has higher throughput?
in the future, the best way to measure relative merits of search engines isn’t quite the relevancy nor the search volume, but quite simply the number of page referrals per session. i would dare say, that we should even add the clicks on the sponsored results to the numerator in that ratio. it would even-out all the other hard-to-measure dimensions out of the equations and offer advertisers the best measure relative throughputs of search engines.