If you use log analysis and if you use a hosted javascript-based web analytics program, you will find that a lot of your hits counts therein may be mutually exclusive.

Let's delve into this issue.

Log analyzers track the information in your web server logs — any time that a web client like a web browser or a search engine spider requests something from your site, it gets recorded to your web server logs.

Due to a number of caching issues, this is not an accurate count of either unique visitors or total page views. Multiple visitors could all be viewing the same copy of your web pages in a proxy server somewhere (AOL being the prime example here), and the same unique user's browser may decline to make new requests for pages that the user has already seen. Log analyzers have a hard enough time divining unique user counts as it is. If a piece of your web content is being drawn from a cache somewhere out there, your web server will simply not know about it. It may be very difficult to try to overcome all caching issues to get an accurate count out of a log analyzer.

Hosted counters that use javascript, on the other hand, can more accurately tell unique users because it can set and get its own cookies and they can count every single page views because they can execute code in a user's browser to force it to notify the tracking program on the user's every move.

However, this is not an accurate count, either for one simple reason: not every client has javascript enabled!

Some people simple browse the web with javascript turned off, and some special browsers do not come with any javascript support. You might expect this, but there is one very important kind of user that do not use javascript: search engine spiders. These kinds of clients will not show up at all!

So which kind of tracking method should you use?

If you want to count impressions for ad views, then use javascript. Because of caching issues, you may be missing hits otherwise.

If you want to keep track of your bandwidth use a log analyzer. Its bandwidth statistics should be exact.

Or, if you want to track hits to non-html files, like images or RSS feeds, then use a log analyzer. Hosted trackers simply have no way of recording these kinds of hits. The same caching problems apply, but there's no other way to overcome the problem.

And by all means, there's nothing stopping you from using both methods at once.