Sep 202010

Several weeks back the guys at NPR music asked a simple question. “What song is the one you’ve listened to most often?” It’s really a simple question, iTunes even automatically maintains a playlist called Top 25 Songs Played. However, I was driven to look deeper.The catch is how do you define listens? Is it per unique tracks on an album, or every instance of a given song across all albums? For instance I have thirty-three “albums” by Dave Matthews Band (including live releases), and as such I have quite a few instances of the song “Ants Marching,” close to twenty as a matter of fact. Now according to iTunes the version of Ants Marching from Live at Red Rocks is different from the Under the Table and Dreaming version, but they remain the same song. So while I could just look at my Top 25 Played list and go with that — Don’t Say “Lazy” by the way — it’s more fun to figure out in aggregate which song I’ve listened to the most. Getting to those numbers requires extracting the raw data out of the iTunes library.xml and performing some analysis.

Fortunately, that happens to be exactly what I do at work, so building a model to extract the bits I wanted from the xml took very little time, though the model was a little sloppy and required a lot of calculated fields to make the data presentable. Summing the play count by song title presented fairly predictable results. As we can see from the chart, Dave Matthews Band songs dominate the top 20.


Plus I’m not even sure how you’d count a song like All Along the Watchtower since it’s technically a Dylan song, but the most plays belong to the DMB cover.

While figuring out that, it occurred to me to figure out which artists had the best ratio of listens per song. Sure, I’ve listened to DMB the most out of all the artists in my collection, but I also have more Dave Matthews Band songs in my library by a wide margin. So it’s really no surprise that they have nearly as many listens as the next three artists combined. The more interesting metric is how many listens per song by an artist?


The first thing that pops to mind when I look at the chart is that the data is skewed by artists who just have a single entry in the library. Take the Boom Boom Setellites, they have the best ratio, but I also only have one song of theirs, Shut Up and Explode (Buy Xam’d now!). I really dig that song, and have listened to it a lot, but that’s not counterbalanced by some piece of crap filler track on the album that I would never listen to. Excluding any artist with just a single song results in what I think is purer data.


There are some surprises in there — for an album I didn’t like I sure listened to Missy Higgins a lot — but on the whole it conforms to what I would’ve said. The Silversun Pickups don’t have many songs in their catalogue, but what’s there is gold. Jenny Owen Youngs also does very well, both on this scale and in number of total listens, which I suppose means if you were to ask what my current favorite artist is, she would be the answer.

Other pieces of miscellany. The average ranking of an album in the library is 59, just a touch below three stars. On average an “album” in my library consists of 13 songs — the number concert albums are offset by all the single song entries. I have not listened to a good third of my iTunes library. That sounds like a lot, and it is, but this current iTunes library is a reconstruction of a previous library, and I don’t really feel like re-listening to all the songs I don’t like the first time. The other metic that would be the most interesting to figure out is what song has the best “played to skipped” ratio, but I’m not sure if my raw data extracted that information or not. It would also be interesting to chart plays over time to see which bands fall in and out of favor.

In the end doing this sort of analysis on my iTunes library is a silly exercise, but it does do a good job of objectively looking at what your musical preferences are. Already I’m trying to think of ways to game the numbers to get certain bands and songs that I think better show off my cred to percolate to the top. But in the end, the numbers won’t lie, and the music that captures my interest will continue to get the heavy plays.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>