Before I move on to the discussion on bandwidth, I want to add a couple more things to the topic of complete downloads that I started in Part 4.
Scott Bourne and Tim Bourquin
provided interesting and relevant responses. They both emphasize that podcasters
have a responsibility not to let advertisers hold podcasting to a
higher standard than other media (i.e. magazines and newspaper) that can’t measure complete content/ad consumption. If they don’t have those numbers, then why should you be required to. In other words, using complete downloads to represent the "reach" of your podcast undersells the value you offer.
The # of unique downloads (including partials) should be an adequate estimate of the "reach" of your podcast. Many of those partials still represent a listeners who engaged with the content and ads.
On the other hand, if you have the resources to measure complete downloads, this metric
provides more detail to the overall picture of your audience and how they are interacting with your podcast. If you don’t have the logs, server processing time or scripts necessary to measure complete downloads, there are a couple ways you can estimate this metric.
Marshall Sponder, who I know from the NYC Podcasting Meetup, is in charge of tracking the podcasting metrics for IBM. On his WebMetricsGuru blog, Marshall offers insight into how he gets a quick estimate of complete downloads:
Another possibility I found gets you close [approximate number of downloads for a file] is taking the number of Hits to a file and multiplying by .80 which seems to be consistent.
Every podcast is usually downloaded a little more than 1x per visitor -
same is true of transcripts of a podcast – that .80 multiplier seems to
work but I can’t tell you why it’s .80 except that I observed this
ratio works several times.
When I take an estimate of the percent of complete downloads for the last 3 episodes of The Podcasting Underground, I get 95%, 75%, 83%, 82% and 73%. This quick samples yields and average of 81.6%–pretty close to Marshall’s figure. Does the 0.8 rule hold true for you?
I got the sbove percentages by looking at the average size of each hit to a file and dividing that by the actual file size. For example, in my web stats (AWStats) I see that the average transfer resulting from each hit to episode #18 of The Podcasting Underground was 15.32 MB. The actual file size is 16 MB. Dividing that out I estimate that 95% of the accesses to episode #18 were complete.
This method was recently used by Rob Walch of Podcast 411 to audit Rocketboom’s metrics for a Business Week article. He took the average size of each download of an episode and compared that to the actual size of the file to get an estimate of how many downloads were complete.
[Rob] looked at the downloads for each day’s show from Rocketboom’s main server and added together the downloads that were done using different kinds of video formats…
Next, he looked at the average file length reported by the server and compared that with the average length of the files at iTunes. This is used to come up with a discounted estimate of what percentage of the downloads were actually completed. (Typically, more than 95% of the show’s download requests were completed.)
I’d like to know what your experience is with these numbers and methods. Please share your ideas in the comments.
December 4, 2006 by Jason
Categories: Metrics