The social news website Reddit has a post “Experts” misunderestimate our traffic, and we dont know why that documents how traffic estimator sites like Compete.com, Quantcast, and Alexa don’t work very well. They compared their own server log data and Google Analytics data to the estimates provided by these other sites.

These traffic estimating tools use their own voodoo to generate numbers so advertisers can compare sites without having access to the actual traffic numbers. The problem is, they’re skewed (and as the Reddit post documents, inaccurate).

For example, Alexa samples users who installed the Alexa toolbar, considered by Symantec and McAfee to be spyware. Do you want to rely on data generated by the kinds of users who would install BonziBuddy?

If you want real usage, you need the server logs. If you can sacrifice some accuracy (resulting from Javascript and privacy blockers), use Google Analytics. If you want a guess, take your pick.