Using the Firefox Browser? Search for "the" using the built in search box that taps into Google, and you'll get about half the results than if you searched Google from its home page. The order of the top results will be the same, but the overall count is dramatically different.
What's Google have against Firefox? Nothing. Instead, I think it's just that for whatever reason, the counts will be decreased if there's ANY client parameter within the search string that shows up when you do a search.
In other words, say you go to the Google home page and search for cars. The URL in your browser's address bar will look like this:
Now let's say you search using Firefox. The URL will look like this:
See the part in bold? The shows you used the Firefox browser client, the client parameter part of the query. Now switch it to anything you like, such as this (I've also removed other parameters that aren't necessary for the search):
Now the "ABC" browser client is being recorded as sending this query -- and the count remains less than that for a Google search without a client parameter.
Note that the issue seems restricted only to when a client parameter appears, not to when a sourceid parameter is included. For example, those using the Google Toolbar within Internet Explorer would see a string similar to this:
That sourceid part shows I've used the Google Toolbar to perform the query. Switch the id to something else, and the count still stays the same.
Want to discuss more or share your own findings? Visit the Why Does The Google Client Parameter Give Less Results? thread in our forums, where inlogicalbearer spotted this unusual situation.
Postscript: Gary tells me that the client parameter brings up the same lower counts that you'll find if you use Google-powered a9 or Amazon results. That supports one theory in this thread over at WebmasterWorld that Barry Schwartz blogged about.
In it, there's a link to an old Google Search Results Protocol page that leaked out, information given to Google partners. It explains that the client parameter is supposed to be used to show which Google partner is calling for the results.
In other words, which client partner of Google -- not which browser client -- is making a request? Firefox developers may have mistakenly made use of the client tag thinking it referred to browser client.
Why would Google's clients/partners hit less than the entire Google database? In the past, search providers have done this to save burden on their servers. Rather than query the entire database, only a portion might be hit -- and that portion might be deemed good enough.
This was the case for much of the Google-Yahoo partnership. A search on Yahoo commonly would come up with half to two-thirds the matches on Google itself. Inktomi used to also operate a Best Of The Web smaller index fed to partners like this, as well.