Spotify, Deezer or YouTube Music have become the main music consumption channel in Spain and much of the world. It’s been a while since we forgot about the CD and the downloads seem to be getting further and further away.
But how do the algorithms of these streaming platforms influence our musical tastes?
Based on data from 4,000 streaming music consumers, Orange sociologists Jean Samuel Beuscart, Samuel Coavoux and Sisley Maillard have studied how these platforms have changed the way the world listens to music and, above all, discovers new artists.
The paper with the conclusions, ‘Music recommendation algorithms and listener agency’, has been published in ‘CAIRN’.
The domain of streaming
Nearly one out of every two euros produced by the music industry worldwide comes from streaming music platforms. According to the latest report from the International Federation of the Phonographic Industry (IFPI), while revenues from physical sales and downloads continued to fall in 2018, those from online platforms grew by 34%.
In fact, streaming revenue ($8.9 billion), driven by improvements in connectivity and variety of offerings, doubled that of the physical sales segment last year. According to IFPI data, 2018 closed with 255 million pay-per-view users. In other words, 255 million people pay to access Spotify Premium or Amazon Music Unlimited, among others.
In Spain, almost three quarters of the music purchased is already purchased over the Internet and there are 2.3 million subscribers to paid services. Of the almost 237 million euros invoiced by the industry in our country, nearly half were entered through payment streaming platforms.
In addition, about one in four euros came through the digital purchase of records and songs and free streaming services financed by advertising.
Algorithms and musical tastes
Radar of new artists, music related to our tastes, custom radio stations and playlists and playlists designed for different moods. Customization of algorithms is part of the raison d’être of digital music platforms.
With tens of millions of songs available for playback anytime, anywhere, to what extent do users give their tastes to the decisions of streaming platforms?
Over five months, Orange sociologists and study authors analyzed more than 17 million reproductions of a total of one million songs. About half of the reproductions reached the end of the song, while 30% did not exceed 30 seconds.
The most popular songs represented only 5% of the total number of songs, but they accounted for a large part of the reproductions.
“The streaming platforms promise that, given their wide offer and their recommendation tools, the diversity of the music consumed should be greater than in the market for physical sales and downloads,” explains Samuel Coavoux, co-author of the study. A premise that was proven in the research, although it varied a lot among users.
According to the study, the most frequent users had a better understanding of how the platforms work and relied more on the recommendation tools and therefore listened to a greater variety of music. However, the more inexperienced users, and especially the younger ones, tended to listen to more famous and well-known artists on a recurring basis.
Even so, 90% of the analyzed reproductions belonged to 10% of the artists reflected in the studio. One in four users exclusively consumed music by famous artists.
“In fact, popular artists dominate both the libraries of infrequent users and the libraries of regular reproductions of active users,” adds Samuel Coavoux. So to what extent did the recommendation algorithms encourage diversity or uniformity of tastes?
In all the analyzed platforms, the algorithms propose songs and artists similar to those listened to, but somewhat less known. However, only a minority end up in the users’ library along with the rest of favorite songs.
In addition, this type of recommendation was consumed, above all, in contexts in which listening to music was not the main activity. For example, background songs for work or exercise.
According to the conclusions of the study, consumption based on algorithms is still a minority. The majority of reproductions are produced through an explicit choice of consumers and not automatically.
Users choose most of their songs and tend to use the recommendation tools only as search engines or to explore discographies of specific artists.
That is, today, personal musical tastes, influenced by other factors, still prevail over algorithmic recommendations in most contexts. However, they have gained weight in some environments where the priority is not to have to actively choose the songs we listen to.