Easily among the more misunderstood concepts about the Internet is the reliability and ability to measure how popular something is. There’s this notion that since one can count hits to one’s web servers, there must exist very accurate measurements of audience metrics. The reality is much more complicated. For example, the web servers are owned by the company running the web site, hardly an impartial bystander in the reporting of audience metrics. Technical issues also make it very difficult to identify humans from automated programs (like search engines). In the end, the methods used for ad rates tend to be a throwback to the way TV ratings are done. While these numbers are considered more reliable than web server stats, they’re not perfect for a number of reasons.

Take a recent example, as MySpace just passed longtime audience leader Yahoo in total pageviews in comScore’s metrics. The announcement has been met with a great deal of skepticism. Ultimately, people who want to live by these numbers, who shift their ad dollars based on who’s “most popular” or “the biggest” need to understand what they’re spending against, and understand the methodology of the measurement firm they choose to consider the final arbiter of these titles. At the end of the day, there really isn’t going to be a 100% accurate count.

Once again, smart people are screwing up audience measurement methods and are complaining that panel-based or survey-based numbers must be wrong because they don’t match their HitBox/SiteMeter/Whatever cookie/javascript based traffic tools. This is just silly. While the panel-based and survey-based reports aren’t perfect, either, their methodology is actually pretty sound. What’s worse, when you think about the numbers, they actually make sense even aside from the math. Fred Wilson points out most of the problems, so read that.

I’ve written about this before, so let me just summarize the two salient points here.

First, automated measurement tools, whether they be fancy, expensive, high end products like HitBox or free/trendy ones like SiteMeter, Google Analytics, or whatever will always, always, always overcount your total visitors. Always. These services can only really track browsers, not actual humans. I browse certain sites from at least 3 computers. Many people will surf from at least 2 (home, office).

Second, real humans are different from visits. Always remember, when you’re looking at your web site statistics, this is the rule of thumb: hits > page views > visits > actual visitors .

The third, tangential point is that Arrington is super silly when he claims that Comscore numbers are “flaky.” Those numbers form the basis of ad rates. If those numbers were truly flaky, which I read to mean wildly inconsistent, no one would use them for advertising rates. Literally billions of dollars are exchanged on the basis of Comscore and Nielsen numbers. Let’s not get carried away with our own self-importance, shall we? In this case, it’s unlikely he knows more than the entire market.