As the US government looks into Google from an antitrust context, a central question has to be what is their real search engine market share in the US? As someone who runs a search engine, I've followed and studied the numbers floating around for a while. And yet they've never really sat right with me.
Comscore and Hitwise are two primary providers of search engine market share numbers. Their two latest reports peg Google's share at around 65% in the US, and that's generally what the press reports.
That seems high indeed, but everyone I talk to "in the wild" who runs high traffic sites actually sees a much higher percentage of their search engine traffic coming from Google, usually from 80-90%. There is some similar data on Quora and HN for reference. If you check those out, you'll see some people (including myself) have seen some sites with less Google % and higher Bing or Yahoo, but these seem isolated cases on particular verticals (and overall smaller traffic sites).
What makes this even weirder is Hitwise also recently came out with a report saying Bing/Yahoo users click on more links than Google users, so you'd expect the Google % everyone sees to be slightly lower than their real market share. But it's the opposite.
Bottom line is something doesn't add up here, and I'm not sure what is going on exactly. Rand Fishkin (CEO, SEOMoz) suggested that it has to do with counting internal links on Microsoft/Yahoo sites:
It's hard to believe this would account for the entire difference, but maybe so. It would be great to see #s calculated off of clicks to external sites. That wouldn't be perfect either and would still overstate Bing/Yahoo as noted above (though that could be controlled for). Hitwise seems to already have that data, so putting out a report like that should be relatively straightforward.
The closest thing I've found to that data is from Statcounter. They report Google's US market share at a much higher 80%, which they glean from their analytic (click) traffic. Their browser share #s don't seem way off, so one might imagine that their search engine #s would be equally representative.
Another possibility is Bing and Yahoo are sending a TON of traffic to a small number of sites. We already know that Bing users use IE way more than Google users, which could indicate that they would be a very different group of people cilcking on different sites.
However, that again seems unlikely to account for all of the difference. You would then expect to see more comments on those Quora and HN threads where people associated with some of those sites would step up and report such.
It's not that Hitwise/Comscore/et al. can't report higher #s. For example, they report 90%+ Google market share in other countries. Just not the US.
So what is it? 65% or 80%? That's a huge difference (~25% in relative terms).
Update: good comments on HN.