Peter Norvig has a good writeup today on how Alexa’s measurement methodology is a classic case of selection bias and bad statistics.

Peter includes some real numbers from various sites and compares them to Alexa’s estimates, what’s interesting is not that Alexa is off, but that it’s widely off even amongst relatively comparable sites.

I can match those sentiments with our own data culled from our logs versus Alexa, Alexa does not track swik.net traffic except in the broadest sense, compete.com is more responsible by not doing day by day numbers when they don’t have the sample or the methodology to back it up.

Really though even bad information is somewhat better than no information at all, and until Google reveals the secrets of the interwebs we’re stuck with a choice between bad and worse.

It doesn’t seem likely though Google will be opening up anytime soon, their lauded Google Trends application they launched as an example of openness hasn’t been updated since November 2006 and cries to open it in the forum are falling on deaf ears there.