Algorithm Update Changes

When you're an outfit like FeedBurner, you tend to see a lot of feed requests from what appears to be a never-ending list of new software clients (350 million requests a day, for those of you counting at home). Maintaining the library of feed-capable clients with names like Rezzibo and @nifty is one thing; interpreting the behaviors of these clients such that we can report subscription statistics is the stuff that makes us FeedBurner.

Today we are rolling out an algorithm update specific to way we report web-based aggregator statistics. Important: This update may result in some FeedBurner publishers seeing an increase in their subscriber numbers beginning tomorrow. Others may not notice any changes at all. For the algorithm enthusiasts out there, we've outlined this change in more detail below.

We live in a world where all web-based aggregators are not created equal. Given that, FeedBurner maintains different algorithms for various types of clients that accommodate different reporting mechanisms. Today, we "retired" one of these - the Maximum algorithm - in favor of its more sophisticated brother - the Cluster algorithm. Sibling rivalry aside, our Cluster algorithm has shown itself to be a more accurate way to count subscriptions, taking into account elusive feed aliases and better reflecting the reality of how many subscribers are accessing your media. To clarify, any increases in subscription numbers aren't "new" subscribers, rather it reflects an improvement to how we calculate data for your feed.

In this crazy world of distributed media, FeedBurner will continue to improve algorithms for existing services while we introduce new metrics that help you understand behavior beyond subscription. We've already started to do that with engagement metrics and Headline Animator stats.