Analyzing CBC Radio 3 Airplay

Late last year one of our songs, Old Hearts, got a fair bit of airplay on CBC Radio 3 (R3). This was really exciting for us as R3 is streamed all over the place, in cafes and offices, over satellite radio, and through the internet to individual users, and we got to experience the thrill of both hearing ourselves and having our friends tell us they heard us on the radio.

One of the neat things about R3 is the built-in web 2.0 features. For any song that’s been uploaded, you can save it to a playlist and play it any time you like, and you can become a fan of an artist and get updates on their activity. If you go to our R3 artist page you can see how many people have added us and how many times they’ve played our songs; when I wrote this article, we had 32 fans, 179 playlist-adds, and a total of 3,424 plays.

When we first uploaded our tracks in May 2010 these numbers were all at zero. Then throughout July they started mysteriously rising. We couldn’t figure out why until we noticed we had been featured on the Music page for the month of July. Intrigued by the information this spike can reveal, and being kind of a geek, I started monitoring the numbers on our artist page, keeping daily records in a spreadsheet, and graphing them. Data was collected through the period of time Old Hearts was in rotation and featured in Grant Lawrence’s podcast, and is current through today, when we seem to have dropped out of the regular rotation. Here are the play counts of the first five songs on our album (Around You, Blue Skies, Old Hearts, On the Radio, World War Z), the entirety of which can be streamed on our R3 page :

Graph showing play count over time

Some really neat things jump out from this graph.

1. The first track gets way more plays than subsequent tracks. The likelihood of a user playing your track decreases as the track gets closer to the bottom of  your artist page. However, over time the more popular tracks will rise up as users discover them and play them more often — notice how the fourth track, On the Radio, outpaces the second track, Blue Skies. Naturally Old Hearts has by far the highest play count as it was the only song to receive airplay.

2. Being featured on the Music page gets you lots of traffic; lots of people will play your tracks.

3. You can tell the level of airplay a band is receiving by the rate of increase in plays per day. It looks like we were in low rotation in August, but shot up to heavy rotation in September following our inclusion in the podcast.

4. Having one song played on the air drives traffic to your R3 artist page. We know this because when the Old Hearts plays pick up, so do the other songs, though not nearly as much.

5. Even after your airplay is done, people will still probably play your songs, presumably because they’ve added you to playlists. Maybe this means you now have “fans”?

Well, let’s find out. This graph shows fan-adds and playlist-adds:

graph showing fan-adds and playlist-adds

What’s likely happening here is that people hear Old Hearts on the radio (through their browser) and then add us as fans and add the song to a playlist, and then maybe explore our artist page and check out some other songs. When the airplay stops, so do the fan and playlist-adds. At the moment we seem to have reached some kind of saturation; playlist-adds and fan-adds are dying down, and plays per day are decelerating.   R3 didn’t tell us when we were taken out of the rotation, but I would guess it was mid-November, given the sudden change in slope around then.

There are two models you could apply to these data to explain the underlying trends: first, music diffusion follows an S-curve, similar to what you would use to model uptake of a new product in the marketplace, indicating that this saturation represents Lakefield reaching its ceiling among R3 listeners. Alternatively, music requires steady airplay to continue receiving listeners; once airplay stops, plays decay. Intuitively is it probably a mixture of these effects, with airplay giving the initial boost until the band’s buzz becomes self-sustaining, as is the case for more popular bands, until listeners eventually saturate; but if there is not enough airplay, a band might sputter and not reach the self-sustaining point.  It’s hard to know which model applies to us — have we reached our ceiling, or did we simply not get enough exposure to reach a critical sustaining mass — since we don’t have enough data. If anyone from R3 is reading this, we would be happy to continue the experiment if you would care to feature another of our songs for a while…

The relatively large base of users and music available to R3 presents some pretty interesting opportunities for analysis.  If one had access to all of CBC R3′s data (which would be lovely) it would be possible to develop an empirical model of the effect of airplay on a band’s popularity. You could control for airplay and try to determine the extent to which intrinsic talent matters; does airplay always help a band, or are there some bands that never catch on? (Conversely, are there some bands that are nothing special, but benefit from heavy airplay?)  You could measure this “talent” with normalized metrics: which bands have the highest playlist-adds per airplay event?  In other words, some bands are so good everyone loves them at first listen; perhaps we can track that and see who these bands are. (Does R3 do this already internally? Maybe this data could be available in the form of a chart? Is this what the R3-30 is? That seems to be more about total plays; perhaps a parallel chart to ‘high response’ tracks).

Now, if you listen to R3 a lot you quickly notice that they draw from a fairly limited heavy-rotation playlist. I think it would be interesting to introduce an element of randomness into their programming – giving airplay to a vast range of stuff, without adding them to the rotation.  Or, you don’t even need to play them on air; since the feedback is available only to internet users, you could simply add a box in the corner of the main R3 page with, say, ten random featured tracks per day.  Then track their playlist-adds following the event. If it spikes much higher than average, it means they have talent: so add them to the full rotation. If not, well, they had their shot. I bet you could tell a hit song and a good band from the reaction after one airplay. It would be great to see some rich number crunching coming out of CBC Radio 3 to take advantage of the opportunities this data provides — and to introduce some more diversity to the airwaves.

2 Responses to “Analyzing CBC Radio 3 Airplay”

  1. [...] This post was mentioned on Twitter by marieclaire shanahan, Lakefield. Lakefield said: The best part of having PhD students in your band? Pointed analysis of your @CBCRadio3 airplay, with graphs! [...]

  2. Ingo says:

    Dude – you guys are legitimately awesome! well done – keep up the great work