My Hit Counter Rant
I remember the excitement I felt as a child when the McDonald's nearest my home would update the sign out front—over 10 million sold, over 11 million sold, over 12 million sold, etc. I remember pulling over by the side of the road as my first car reached 100,000 miles, and the odometer rolled over to display all zeros. More recently, I've seen hit counters on pages throughout the web.
People like numbers. We like pointing to numbers and saying, "See, it must be good." The numbers don't even have to mean anything. Those 10 million other hamburgers didn't make mine taste any better. My car certainly wasn't new just because the odometer displayed all zeros, and I hadn't even driven it for most of its 100,000 miles. A web page doesn't become useful (or even "kewl") because its hit counter has been incremented some number of times.
In the case of hit counters, the absurdity is compounded by the inaccuracy of the number displayed. At least the number displayed by the sign at McDonald's or by my car's odometer might be reasonably accurate. Hit counters are inherently inaccurate.
Then things get more absurd. Web authors discover that many people surf the web without loading images automatically, and so they include alternative text (an ALT attribute) along these lines: "Make your visit count. Download this image." Are my visits irrelevant unless I increment their hit counters? This reminds me of those TV ads that tell viewers to call some 900 number to vote on whether they think O.J. did it (or whatever). Why should I pay for the privilege of helping you collect meaningless data?
So why are hit counters inaccurate? There are a number of reasons, actually. As I mentioned earlier, many people surf the web without loading images automatically, and these hits are ignored by the most common hit counters—those that use inline images generated by a CGI script. Many browsers cache data, and hit counters ignore retrievals of these cached copies. More and more users (e.g., AOL and WebTV customers, and many European users) are using cache proxy servers, which retrieve the page once and then deliver it to any other user who requests the page without retrieving a new copy. (I expect cache proxy servers to become even more common as web users and providers respond to the increased traffic and longer delays on the web.)
If you want statistics about web server activity, learn to read your server's log files (ask your system administrator if you need help). If you want to know how effective your web site is, collect feedback from your readers. Your hit rate (regardless of how you determine it) reflects your site's publicity, not its effectiveness.