How can we tell how many people subscribe to our podcast feeds? Unfortunately, there is no central registry of subscription information. Few feed aggregators and directories announce how many people have subscribed to each of the feeds they index. One example I could find was the honest and open people at Bloglines, always one of my favorite feed organizing tools. A second is Google--which sticks the number of Feedfetcher subscribers it represents into its agent request string. It is possible to get more info by using Feedburner statistics. Unfortunately, with our dynamic approach to feed creation, this is tricky to implement.
So, what can we do, on our own? We know that 844 distinct IP addresses downloaded our RSS documents, on January 16th. 57 of them requested more than one of our documents, for a total of 901 IP address/document combinations. Of these 901 requests, 530 were made only once within twenty four hours. It turns out that most feed readers automatically check a subscribed feed more than once per day. (The default for iTunes I believe, is to check a subscribed feed every six hours.) So, we can guess that the 530 single requests out our total of 3,113 RSS document requests were NOT from subscribers. In fact, we can probably exclude as a subscriber anyone who asked for our document fewer than every six hours or five times in a day. This allows us to eliminate 760 of our IP address/document combinations and 1,109 of our RSS document requests.
We now know that our maximum number of subscribers would be 3,113 - 1,109 = 2,004 RSS document requests / 4 = 501 subscribers. Remember that we had one example from the feed reader side--the place where the subscription records are kept. We had 40 requests for our RSS documents from IP address 65.214.44.29, an IP address that reverse DNS shows is Bloglines. They claim they handle three subscribers for our podcasts. Bloglines check feeds for their subscribers, once every hour. That means that our three subscribers ought to have generated 3 x 24 = 72 requests for our RSS document. We found 40 of these. I suspect that there is another IP address for Bloglines (I didn't reverse DNS them all) that has the other 32 requests.
We can also get an estimate of subscribers from our file download log. We happened to post new episodes on each of our three house feeds on January 16. In theory, the podcatchers of those users who subscribe to our feeds should see these new episodes and download them within a relatively short time. It turns out that 395 of our 2314 downloads occurred within twelve hours of their time of release and 231 occurred within six hours. These numbers are in the same general ball park as the 501 number above. I feel comfortable saying that between 10% and 15% of our downloads were from subscribers to our podcasts.
Around 1,454 listeners downloaded our files.
554 users came from podcast directories and other sources. Some of these may have subscribed to our feed, after trying it out.
About 900 users learned about us by browsing our RSS document. Many of these downloaded our podcast files almost immediately afterwards.
Between 230 and 345 of these immediate downloaders were probably subscribers, who were using a podcatcher to monitor our feed.
This creates a picture of a core group of dedicated listeners (300?), a group of regular visitors who "check out" our RSS feed on a directory and pick up a file when it interests them (600?), and a group of random visitors who pick up a file when they hear about it (550?). You can see that it takes some work to accumulate subscribers and that "casual" visitors are important to overall file download numbers. The best way to boost subscribers is probably to regularly add new (and interesting) episoides. The best way to boost casual visits is probably to improve your search engine optimization and to upload to multiple podcast directories. We'll talk about both of these issues in future posts.
In commercial terms, we can claim to reach between 10,000 listeners and 25,000 listeners per month. If someone paid $10 for a "CPM" (Cost Per Thousand) worth of advertising on our site, and we described our site as having 2,000 downloads per day x 30 days (how it is normally done) = 60,000 downloads a month, they would pay us 2,000/1000 x 30 days x $30 = 1,800 for one month's advertising. Our media would cost $1,800 / 25,000 distinct listeners = $0.072 per ad listened to. Don't shudder--radio and TV can be much more costly than this, when you measure actual cost to reach a listener! Since we download new podcasts each week, our 300+ core listeners would get up to four impressions during that one month period--a useful fact that makes our feed a bit more valuable in advertising terms.
I hope these posts have been helpful. Let me know if you have ideas or tricks for understanding podcast statistics that I have missed.


