Variety in Public Radio Music:
A Look at the Data.
On January 24th, 2005 a new radio station came on the air in Minnesota. Broadcasting on 89.3 FM, The Current The Current promised to bring an interesting variety of music to the Twin Cities, and to highlight local artists. From it’s first song, Shh by Atmosphere, The Current fulfilled this promise.
It was only a few months after The Current started broadcasting that I was introduced to the station. A friend I was working with in the Summer of 2005 started turning on The Current while we were working. I was quickly hooked. At that point in my life I knew that I enjoyed finding new music, but I hadn’t found a source that constantly provided me with good music. The Current became that source.
I wasn’t the only one listening either. Readers of the Minneapolis newspaper, City Pages, voted The Current as the Readers Choice Best Radio Station of 2005 (and of 2006, and of each year thereafter). By 2008 The Current won the award proper.
When I left Minnesota to go to school in Portland, Oregon, I missed The Current, and I always looked forward to hearing new music when I would come home for breaks. It seemed, though, that every time that I came home, the music The Current was playing was a little more generic, and the playlist was more repetitious. I still listened, and still heard new music on the station, but I didn’t like hearing a few of the same generic indie-rock songs over and over again.
I got fed up, not long ago, when I turned on The Current on a Thursday night to hear their weekly program
The Charts Show. Ostensibly, program plays the top 20 songs for the week, as voted on by the listeners. While I don’t believe the voting is rigged, it always seems that the songs that are in the top 20 are also the songs I’ve been hearing far too often for the past week already.
Fueled by the frustration of having heard Ho Hey by The Lumineers one two many times, I went to the online playlist and scraped data for all of the songs that have been played on The Current between January 1st, 2006 and December 31st, 2012. I had to know: Has The Current really be getting worse? Or am I just turning into a radio grouch?
The results are interesting. In the seven years from which I collected playlist data, The Current played 805,604 songs. This averages out to about 13.1 songs per hour over the course of those seven years. About 5% of the data collected contains errors or missing values. Once we clean and normalize the data set, we end up with 760,146 plays to analyze.
The simplest way to check if variety is declining over time would be to check the number of different songs that are played each year. If The Current is drawing from a smaller set of songs, it would be fairly clear that variety is decreasing. The following table shows the total number of songs played each year (including repeats), and the number of unique songs played each year.
|Year||Total Songs Played||Unique Songs Played|
This table doesn’t go very far towards answering our question. There seems to be a dip in the number of unique songs played in the year 2008, but that number increases from there on. Also, the total number of songs played seems to be increasing as well.
This table doesn’t necessarily reflect what you would hear if you were listening to The Current on a regular basis. Most of those unique songs could be played only once in the year while just a few songs get played over and over again. A better measure might be how much time is spent playing the 50 most played songs. Or maybe the 100 or 1000 most played songs. Figure 1 plots all of these (and every step in between) for each year between 2006 and 2012.
It is immediately apparent from Figure 1 that 2006 and 2007 are different than any of the other years. In 2006, If you were to listen to every song The Current played, and mark down every time you heard one of the songs that ended up being in the top 50 songs that year, you would only make a mark for 2.7% of the songs you heard, whereas in 2012 you would be making a mark for 10% of the songs. If you were to do the same exercise for the top 250 songs you would mark down 10.7% of songs in 2006 and 34.1% of songs in 2012.
Taking this analysis to the extreme, in 2008, the top 1,000 songs represented 61% of airtime. Looking back at the table above, we can see that 10,518 unique songs were played in 2008. The top 1,000 songs, then, represent just under 10% of songs The Current is willing to air, but they spend 61% of airtime on those songs.
To a large extent, though, it’s not the 1,000th most played song that would make the station feel repetitive. Throughout the course of a whole year, the 1,000th most played song doesn’t go above 25 total plays in any year in the data set. What makes a station feel repetitive is when you hear one song over and over again. At that point it doesn’t have to be over the course of a whole year. In fact, it doesn’t have to take more than a few days to notice that a radio station is giving a disproportionate amount of airtime to a particular song.
To get at the effect of hearing one song over and over, we can look at the number of plays the top-played song in a given week gets. This is plotted in Figure 2 below. This view also has the benefit of giving us a fine grained look into changes in the station over time.
Looking at Figure 2, the most striking feature of the graph is the jump that occurs right around week 104 (the start of 2008). The Current went from playing their weekly top-song 4-7 times a week to playing their weekly top-song 15-20 times a week. This obviously a numerically significant difference, but it is a qualitative difference as well. When you are significantly above seven plays a week you’re in the realm of hearing the same song multiple times in a single day.
After noticing the significant change at the end of 2007, it’s also worth noting that the number of top-song plays peaks at 21 plays per week at the beginning of 2009, and then declines back to around 16 or 17 plays per day. Interestingly, it seems that the three dips in Figure 2, at the end of 2010, 2011, and 2012 correspond to a time when The Current is playing more holiday music.
So, how and why did this happen? Figure 2 shows an abrupt shift in how often the top songs get played. It seems unlikely that the DJs at The Current suddenly decided that they wanted to be playing the same songs over and over again.
After doing some digging I found that in late 2007 the current hired a consulting company in response to declining numbers of listeners. The suggested measure was to curate a set of songs to play in heavy rotation at any given time. City Pages reported on this in 2008, describing the controversial decision give a
new mandate to DJs that
puts a handful of new indie hits in heavy rotation.
This change fits quite cleanly with the data. The change to a heavy rotation model can be seen in the large difference in the proportion of airtime taken up by the top songs in 2008 as opposed to 2006 or 2007, and the time of the shift itself can be fairly closely pinpointed to the first few weeks of 2008 by looking at Figure 2.
The natural question to ask next is whether or not the change had the desired effect. Did the ratings and regular listener count go up? According to MinnPost
In December 2009, a mere 1.5 percent of local radio listeners tuned in to Minnesota Public Radio’s The Current […] by June, that 1.5 had nearly tripled to 4.3. This would suggest that the early 2008 change to a heavy rotation playlist did not, by itself, cause the ratings increase.
The same article by MinnPost would argue that a some portion of the ratings increase seen in early 2010 can be attributed to the 2009 hire of Jim McGuinn as the Program Director for The Current.
McGuinn hit town after the Current’s playlist underwent a major upheaval — from anantiformatwhere DJs literally brought in favorite CDs from their collections, to a tighter, consultant-advised format that caused one part-time host to quit over the moroseness.
It was a fairly abrupt shift,concedes morning co-host Steve Seel, noting the playlist becamemuch tighterbefore McGuinnrelaxedit.
This slightly more relaxed approach, along with careful tracking of playlist data by The Current itself seems to be working for increased ratings. It may be that this policy change by Jim McGuinn is the cause of the slight decrease in number of plays of the weekly top song that we see in Figure 2 after 2009.
My personal conclusion from this investigation is that, while I may not always be pleased that The Current has chosen a model that emphasizes fewer tracks at a given time, there are clearly benefits to the station, and I would like the station to survive. I’ll still listen, and they will still introduce me to new music. It is certainly fun to be able to confirm a number of suspicions about the quality of their playlists, and this is a cool starting point for a number of directions of inquiry.
Despite the decline in variety over the years, if you’re looking for a good list of really good indie/alternative songs, this list, produced from this data set, of every song that has been the top played song on The Current in a given week is a great place to look. Or, a slightly more curated list would be the top ten songs of each year.
The code used to do this analysis can be found in this gist.
- Comparing The Current’s playlist to those of other radio stations, both radio stations that target similar listeners and radio stations that play differnt music (top 40 stations, country music stations, classical music, etc.). Further, can we combine this data with actual listener data?
- Look at the effects on The Current’s Chart Show. How fair is the chart show? That is, are the choices people make influenced by playcount? Even if we control for the quality of the song?
- Analysis of the factors that go in to choosing songs for heavy rotation. Using a music database (such as (http://www.discogs.com)Discogs) to pull in more information about these songs, does a factor like track length have any bearing on the likelihood of a song making it to heavy rotation? What about independent listener ratings?
- Is it possible to refine the variety calculation to include hourly trends in listenership? That is, should we weight songs by how many people hear them (e.g. a song played every day during drive time drive time is more likely to be heard by more people)? How does this change the trend over time?
Any questions, comments, suggestions about this post? Or anything else? Feel free to send me an email: TheErikSwanson@gmail.com.