The Blogger guys have added really simple-to-use statistics gathering in their "Blogger in Draft" beta page, so I've been trying it, comparing it to StatCounter and Google Analytics, both of which I've embedded in Perils of Parallel.
The interesting thing is that all three give different numbers. For example, Google Analytics always gives smaller hit counts than StatCounter, and both are lower than Blogger Stats. So obviously I prefer Blogger Stats. The differences are major, like a factor of two at each stage (so Blogger Stats is 4X Google Analytics).
I guess it's not horribly unexpected that there are hit count differences, although those magnitudes are a bit worrying. But I wouldn't have expected such differences in their capturing what browsers and OSs are used.
All three say Firefox is #1: Analytics says 49%, Stats says 42%, and StatCounter says 46%.
But Analytics and StatCounter say Chrome is #2, at 25% and 23% respectively; while Stats says #2 is IE, at 27%. #3 for Analytics & StatCounter is IE, at 9% and 15% respectively, with Stats saying Chrome is #3 at 13%.
I was thinking the difference is in duration over which the numbers are gathered, but it doesn't change much between last week and last month.
There's a similar inversion in operating systems: Windows is #1 all around, but StatCounter says MacOSX is #2 while Analytics & Stats give #2 to Linux. (Stats does so only indirectly, just saying the "OS" is X11; I wonder how many people stare at that wondering what X11 is.)
So, my take is that nothing's very reliable. But it's fun to fool around with. I'd be more concerned if I were getting any significant revenue from AdSense, since it's based on Analytics. But I'm not.