In the 1990s, the global music industry was riding high on the back of huge CD sales. Online piracy and technical innovation led to a tumultuous period for music that ultimately led to the current streaming age. My project uses data to examine how these changes have impacted the economics of the music industry, and how it’s reflected in the artistic choices of musicians and producers. One of my key insights is that the payment structure of streaming led artists to make shorter songs. This is a scoop of analysis that started with reading an analysis by the data scientist
Streaming now accounts for an estimated 80% of all music industry revenue. Prior to this reporting, how that shift is affecting music itself was little noticed or understood. Plus, many music consumers are unaware of how music streaming payment structures work. Through data analysis and compelling graphics, I was able to write popular articles that helped people understand why music may be evolving differently in the streaming age. Our project points out that artists receive payment for songs after 30 seconds of listening, and that an artist receives no more money for a 5-minute song than a 2-minute one. I also point out that streaming revenue is distributed using a “pro-rata” system rather than a “user-centric” one (see the article “Your Spotify and Apple Music subscriptions pay artists you never listen to” for a technical description of the distinction). This payment structure disproportionately harms classical and jazz artists.
Our work on how streaming is changing music has been widely covered. It was made into episodes of the podcast “Switched on Pop” and the Austrian radio show “The Money.” It was also covered on NPR and the BBC.
For this project I used the statistical programming language R, Adobe Illustrator, and the chart-making tool The Atlas.
I went to find data relevant to my hypotheses to analyze. I used R to collect data on the lengths of songs from the Billboard Hot 100 and the Spotify API. Once the data made emerging trends clear, I then took this data and used the R charting library “ggplot2” to make basic versions of the visualizations I would use to show how particular artist’s songs were getting shorter over their subsequent albums. I also used R’s data analysis library “dplyr” to get summary statistics of the length of songs in each year, and how that differed across genres. To collect data from the Spotify API, I used the library “SpotifyR.”
After making basic versions of the charts in ggplot2, I took those charts and cleaned and annotated them in Adobe Illustrator. I also made several different versions of the visuals in Adobe Illustrator so that the charts looked good on different screen sizes.
For simple line charts and bar charts, I used Quartz’s data visualization tool The Atlas.
What was the hardest part of this project?
The most difficult part of this project was coming up with a compelling way to demonstrate how streaming economics actually impacts music listeners’ lives. It was very simple to show that more revenue now comes from streaming and songs are getting shorter. A simple line chart or bar chart suffices to show the trend.
But to drive these changes home and ensure they really resonated with readers, I decided to look at particular artists and how their albums changed over time. This involved thinking creatively about how to show the composition of each album in terms of song length. I chose to use a horizontal stacked bar chart in which each section represented a song. This approach would not necessarily be intuitive for all readers, so I put a lot of thought into how to annotate and color the chart in ways that would help make the visualizations more approachable.
In addition, I made visuals for the last three albums of seven different artists, which took time and care. I felt that seeing this pattern for such a large number of artists would help bring the point home.
The result is what grew into an engaging project consisting of multiple stories, each one telling a different facet of the larger narrative about artists, music companies, consumers, and platforms, all grounded in and powered by data journalism.
What can others learn from this project?
Data journalism can sometimes suffer from being impersonal. Typically, data analyses are looking at averages of large groups—such as people, countries, or, in this case, songs—and finding a trend. These trends often tell us something important about the world, but by their nature, they are abstractions. One way to deal with this problem is by interviewing people who are examples of the trend or who are impacted by it. I tried another tactic for this project, which I think others can learn from.
This project was successful because I started with a thesis and then took a broad finding, “songs are getting shorter in the streaming age,” and then honed in on specific examples. This helped readers connect to the reporting. By looking at the last three albums of popular musicians, for example, they could see how streaming changed the music of artists they love and the music they listen to and buy. I believe that this is a good lesson for data journalists. Humanizing the data through specific data examples can make all the difference.