While the new google reader stats are a neat idea, Mashable comes up with a hugely researched piece on why the new Google stats are not entirely accurate. With enough supporting information to make you wonder just how accurate any web based reading system can be against the logs that the server itself is generating.
Maybe it is time for people to cough up some logs and see what is really happening with the delay between actual log parsing and on line statistics for items on the web server. For example my on line tracking will show 100 visitors while my direct log parsing system will show 300 visitors, consistently logging in 1/3rd on average all the visitors to the web site..
It looks like feed burner and the new Google stats are looking at the same issue. In that the numbers are not going to be accurate, and not accurate depending on where you are in their system, and how they read/rank the blog or site you are working from.
A default feed, in case you don’t know, is a feed which is presented to users on signup. Google Reader, for instance, pushes new users to these feed bundles: instead of searching for feeds you like, just grab a bundle on a certain topic. This is a great boost for those sites that can get themselves listed in these bundles: often by striking a deal with the feedreader company or being friends with the owner. Source: Mashable
Obviously being part of a random bundle will do someone well as long as the person or the company can afford to be part of the bundle. With a lot of default software, just because it is in a default bundle does not mean that people are actually reading your feed. In office 2007 outlook you get default feeds for Microsoft at Home and Microsoft at work. Many do not delete these feeds because something interesting might cross them someday. This really jacks up the numbers of readers to a site.
It also jacks up the stats to the page call on the web site. Of the 3000 some people that came to the site, some 1666 of them are right to the PHP based feed that one site uses. But that is not reflected in the statistics that are presented by the on line subscribers numbers from feed burner.
At this point, most of the web based statistics makers are going to be off, and in what Mashable is saying, they are manipulatable as well via bundling. So depending on what you want to do with your feeds, and how much you need to show a major amount of traffic, maybe getting in on a bundle of feeds is not a bad way to go. Any information though is going to be dubious in the longer run, because with out direct log reporting, there is really going to be no way to prove that those feeds actually still exist.
The key to the feed numbers is to make sure that they are updated daily, which feed burner does, and page flakes did not do. There are still ways around the system though, and in a world of black hat SEO, and splog web sites, this would be one way to jack numbers and get a better advertisers bid rate.
