jasoncalacanis’s posterous

Stats are pretty much lies all around the board.

2. we use Google Analytics and ComScore and I would say they are within 2-10% (mainly because we load QuantCast last on the page and Analytics earlier since we get more from it).

sample based data = bad.
real tracking pixel data = good

that simple

No I definitely don't work at comscore or any other analytics company. I've just dealt with them all for a very long time and understand how they work behind the scenes. I wouldn't put a whole lot of faith in any of the numbers anyone gives you.

It only has a US panel, so international traffic is ignored
Panels skew heavily towards home users, so work traffic is generally underrepresented
It's sample based and extrapolations are always a guesstimate, particularly where low volumes limit statistical significance.
- These issues alone would easily explain the PV variances above.

Some reasons why Quantcast struggles:

I'm not familiar with Quantcast, but it seems to use the same tag based method as most Web Analytics vendors, so it is essentially measuring cookies and javascript - it isn't perfect, but it's about as good as it gets for most sites. Page views should be relatively accurate (compared to unique visitors for example)

I'm a bit dubious about their claims re measuring demographic and lifestyle etc. Any JavaScript based method won't have this so it will presumably be extrapolation based, so subject to the same issues as above.

IMO, it's that they post their numbers publicly.