New Detailed Statistics Page

In the few months since we released the first Pre-Alpha of FeedBurner (more on "pre-alpha" later), we have begun to manage feeds for thousands of publishers, and we've learned an incredible amount about the behavior, characteristics, and nature of the various feed readers and online aggregators out there.

Today, we are introducing a number of new measurements on the detailed statistics page in an effort to give you a much better window into how your feed is being used. First, we have added the notion of "Daily Circulation". Circulation is an approximation of the number of people on whose behalf your feed was accessed today. Ok, so what does THAT mean? Generally speaking, there are two types of aggregator that might access your feed: feed readers that run as software accessing the feed on behalf of one person (eg, FeedDemon), and online web-based aggregators that access the feed once on behalf of many people (eg, My Yahoo, Bloglines). There are also two types of non-aggregators that might access your feed: browsers and bots. Your feed registers a browser hit whenever somebody visiting your blog in Internet Explorer clicks on the little xml icon you've got, for example. Bots are the various search spiders and crawlers out there indexing all the files they come across on behalf of search engines and the like. We call anything a bot that is not accessing your feed on behalf of any person(s). The Google crawler, Googlebot, is one example. Having said all that, daily circulation is, technically speaking:

Unique IP addresses accessing your feed today on behalf of a unique client feed reader plus the number of subscribers to your feed reported today in any accesses of your feed by an online aggregator.

Whether a unique IP/client combination accesses the feed once or two thousand times in a day, that counts as 1 toward your circulation. I.e., circulation is an approximation of the number of subscribers to whom your feed has been distributed today.

Browser hits and bot hits do not count toward circulation. That may raise another question for some of you who are quite savvy about the variety of bots and browser extensions that serve as feed readers out there. We have catalogued hundreds of user-agents and worked to determine which are readers, which are online aggregators (there are MANY emergent online aggregators around the world), which are browser hits, which are feed reader plugins to browsers, and which are bots. We have by no means catalogued everything yet, but we've made a pretty good first pass.

We are excited by the results we've seen in testing. We think circulation presents a far better view of your subscriber trends and daily usage than the previously discussed notion of "new visitors." You will also now see your click-throughs on a daily basis over time, so that will provide some wanted detail that many of our publishers requested. Be sure to look at your detailed circulation numbers, in which we break down circulation by feed reader and online aggregator as this will start to give you a better view of the "hits" to "circulation" discrepancy that's a bit unique to syndication.

Circulation is by no means perfect. There are a few different issues with trying to make a highly accurate count via this approach, including the usual IP address as person issues (although we try to mitigate this issue by only using "daily circulation" and not "weekly" or more, so that you have less likelihood of DHCP users reporting multiple IP addresses over the time period). The key is to think of circulation in much the way, well, commercial publishers think of it. It represents the best current approximation of how many people you reached today, via the various agents reporting back to us through feed accesses. This number is particularly interesting as a trend over time.

This post is getting quite long as it is, so just one last note on circulation accuracy: some online aggregators do not report in their user-agent how many people are subscribing to your feed through that online aggregator. On your new detailed stats page, we highlight any of these online aggregators that are accessing your feed (using the following icon: ) so that you'll know when your circulation is underrepresented by some unknown number.

Are there mechanisms by which you could make a highly accurate count of number of subscribers, which items were read, etc. ? There have been a few suggestions made, but all of them have issues. Briefly, the suggestions range from a couple of web world mechanisms like cookies and the transparent GIF, to a more syndication specific notion of every subscriber getting a unique URL for the feed. We'll probably post on the subject soon because we've thought about it a lot (e.g., one problem with unique URL is that it pretty much breaks the simple notion of sharing your OPML). Let us know what you think of your new detailed statistics and stay tuned as we begin to roll out some entirely new services.