wow, that is a lot higher than most would expect!

That is actually a small number since I think he speaks about distincts. Also it quite fits Pareto 80-20 rule.

This isn't saying that the 80% is trivial - it's just a stat on handling user uniqueness. So for me, 20-25% doesn't shock me at all.

Well some stats have offered stuff like 50% were unique to that day, but with X billion searches I probably would have expected more overlap than 20 to 25% never being searched even once on Google.

Just check your access logs or your analytics: more people are doing multiword, long searches than they used to, so that factors into it.

"that guy on oprah the other day talking about dog training"

I take this takes into account things such as spelling errors, gramatical errors, joining of words when they shouldn't be, because in that case I'd believe it! Another things is searches in other languages such as Italian have different spellings for different words i.e. different language dialects.

This made me feel good.

(But I really hope it's not just misspellings.)

Jim Lanzone (Ask) said:

On any given day, 60 percent of the search requests we get, we have never seen before.

http://www.msnbc.msn.com/id/18947235/

Is that a fair reflection of the comprehensiveness of their respective index databases? :P