It has been interesting over the last few days to watch the various statistical reports that my hosting provider provides coming online and providing conflicting reports.
It seems that none of them can agree on how much bandwidth the site has used (it ranges from about 5MB to 15MB) and they all seem to want to tell me different things about how many 404 errors there have been.
The most interesting statistical anomaly so far is that AwStats is estimating that 295% of the readership have added the site to favourites. If it is true that, on average, each reader has added the site to favourites 2.95 times then I am very pleased, but I suspect it is just AwStats having problems with it’s estimation calculations.
I am still continuing to use the offsite statistics collected via javascript and an image (Webstat & Nedstat) for the moment. Whilst they don’t provide the purely accurate statistics from the webserver logs, they do measure page views pretty well, and are much better at counting visitor numbers due to the use of cookies. (Don’t worry, I have checked the privacy policy of these companies and they don’t even look at the data they collect.)
I will probably download the raw logfiles and analyze them myself with something like Pyflag [1] which is (at least partially) the brainchild of Michael Cohen, organiser of the Security Miniconf [2] at the recent Linux.Conf.Au [3]. PyFlag will provide me with the power to find out what I want to find out accurately, which is more than I can say for the other statistics in use which do their best, but aren’t flexible enough….plus I get to have more fun this way.
Samuel