There are several issues with their methodologies. Basically, they measure data from toolbar installs, as well as some data sharing/sampling agreements with IS

There are several issues with their methodologies. Basically, they measure data from toolbar installs, as well as some data sharing/sampling agreements with ISPs and other websites. Most criticisms center around the fact that it is well documented that data from compete frequently doesn't match (or come close to matching) data in server logs. Because of the methodology of collection, the data has a tendency to be skewed. Plus, it's difficult to figure out exactly why server logs show such different data than compete, so it gives a not-too-trustworthy feel to many webmasters. In addition, there's a potential conflict of interest -- if you sign up for their paid service, your site will get an analytics tracking cookie. From what I've read, people rarely see their rank drop after signing up for the analytics service. Gives the whole thing the feel of payola. That said, similar problems exist with all the analytics type services. None are perfect, best bet is to poll a few of them and not take the numbers too seriously (look for trends, not milestones).