Not that I personally need to know how to extract fairly accurate UA-ratios since I'm only interested in visiting/using-trends - regardless of UAs. However, it seems like it is possible to get some "quirks" in those trends too, so how big a "pinch of salt" should one add just to be (somewhat) on the safe side?
A fairly generous pinch, I think :) Realistically the best you can do with current technology is keep a watch on your own stats and watch for the fluctuations within that realm. You have to keep in mind some known oddities in UA identification too, as well as any known influences - eg. large organisations may lock down their users to a specific browser, which will increase that browser's share. In terms of usage patterns, perhaps the biggest unknown is UAs' caching/pre-caching habits. As in, a browser which is extremely aggressive in its pre-caching habits will deliver more hits to the server than the user actually created; or a page may be reloaded from a very frugal caching browser so two hits/views are counted only as one. I talk a bit further at http://weblog.200ok.com.au/2005/06/lies-damn-lies-and-browser-statistics.html although really the main points are covered here. cheers, Ben -- --- <http://www.200ok.com.au/> --- The future has arrived; it's just not --- evenly distributed. - William Gibson ****************************************************** The discussion list for http://webstandardsgroup.org/ See http://webstandardsgroup.org/mail/guidelines.cfm for some hints on posting to the list & getting help ******************************************************
