Posted 25 May 2008 - 12:17 PM
The fact is that all analytics is based on making some very subjective decisions. How long between one page view and another by a single user should count as a separate visit, for instance? 30 minutes? An hour? What makes a visit unique? If a second machine (different user-agent) from the same IP address (a shared line) a unique? The guy who arrives at your site on the PC at work during his morning coffee break, then returns that evening on his home PC is certainly going to be tracked as a Unique, not a returning visitor, by any off-the shelf analytics package.
Then there's how the cookies work. Some cookie tracking is more commonly blocked by anti-spyware than others. The duration of the cookie may vary too. Then there's how you define unique: is someone unique to the current sample of data (i.e. was unique in that week, but was there six weeks ago) a Unique, or a returning visit? Choices by analytics vendors vary.
Next there are more technical issues. Analytics based on a bit of code in the page and a cookie is in many ways less reliable than server-based solutions. i.e. to get a page from my server, my server must recieve (and ths log) the request, where a third-party tracking solution, such as Google Analytics can be blocked all too easily. But what about return visits or multiple page-views where browser caching means yor browser still has the page stored and doesn't request it from the server? Sometimes there the third-party solution may be the more accurate.
They won't ever agree.
But none are ever actually accurate at all anyway. They are unly a rough guideline and good for detecting trends.
So just pick the one you feel you understand the most, and use that. You're looking for the trends, so need a set yardstick, even if it is actually a metre stick, or a 30 inch stick. So long as you use the same solution from month to month, then the trends will be correctly detectable to the limits of the accuracy of tracking anyway.