In Podcasting Metrics Series Part 2, I pointed out the problem with looking at the # of hits to a file for estimating the # of downloads. A single download can result in many hits. The hits stat will always be bigger than the actual number of downloads.
So how can we get a more accurate idea of the number of downloads? Services that offer podcast metrics (e.g. Libsyn, Podpress, Feedburner, Podtrac, RadioTail Ripple etc.) track things such as the IP address of each request to a file to determine how many unique users have downloaded (or at least attempted to download) the podcast file. This leads to a more accurate, although still imperfect, measurement.
This is probably the most widely used method for estimating audience size. But is this sufficient for advertisers?
The problem is the numbers can still vary from platform to platform
(and hence podcast to podcast) depending on the algorithm used to filter the logs. This makes it difficult for an advertiser to compare show A to show B when deciding who to advertise with. At the same time, I have found the unique download numbers to be relatively close between Libsyn, FeedBurner and Podpress.
At the very least, measuring unique downloads gives you a relative look at the growth trends in your audience. You can see that episode X got listened to 20% more than episode Y and draw some conclusions about audience growth and/or popularity of topic. It’s a metric that I check regularly using both Libsyn’s stats (they host some of my media) and Podpress (the plugin I use to publish my podcast episodes to my blog).
Another shortcoming of the unique IP download metric is the fact that a single listener could download the show at home and at work (two different IP addresses) hence still being counted twice.
In addition partial downloads are still counted. Someone could have stop grabbing the file after streaming only 30 seconds. At the same time, a partial download could be someone who listened to 80 or 90% of the show. Do they both count as listeners?
How can we measure complete downloads? Is it important to measure complete downloads? Further, we still don’t know if the file was listened to after it was downloaded.
As one final comment I want to point out that, although these numbers have their place, Michael Geoghegan has shared on multiple occasions that Grape Radio (who charge a healthy ad fee per episode) have never been asked for download numbers from advertisers. I’ll be discussing this as well.
In part 4 I’ll continue the discussion about complete download measurement. Then we’ll get into subscriber measurement. It looks like this series will continue for awhile yet. I’ll take it as long as it needs to be. Please share your thoughts and questions in the comments.
UPDATE: There’s one more thing I should point out about measuring unique downloads. To get more accurate results, it’s important to filter out hits from directory bots that are checking for and indexing new content. These are automated services and not actual people consuming content. Each of the metric services I have mentioned do this a little differently.
Email: jason AT jasonvanorden.com
Voicemail Line: +1 877-877-7799
0 Comments For This Post
2 Trackbacks For This Post
November 29th, 2006 at 3:31 pm
[...] Jason Van Orden is taking a crack at the podcasting bugaboo: Metrics. 3 parts so far, (1,2 and 3.) [...]
December 1st, 2006 at 1:26 am
[...] In part 3 of this series on metrics, I stated that measuring unique downloads, while still an imperfect podcast metric, was one of the most popular ways to measure a podcast’s audience size. While filtering your web logs by unique IP cuts out multiple hits from one user, you still don’t know if that user received all or just part of the file. Measuring downloads that are not only unique but also complete results in a more conservative metric than measuring just unique downloads. [...]