Exploring Local Music’s Place in Global Streaming
May 21, 2024 Published by Samuel Way, Dominika Mazur, Rebecca Kupferman, Megan Walsh, Angelina Tizé and Ben Lacker
tldr – Music has always been intertwined with place and culture. Now that digital streaming has made it easier than ever to encounter music from all over the world, what role does geography play in people’s relationship to music? We use a mixed-methods study to examine the definition of “local” music and explore its application to music recommendations. Our qualitative and quantitative results make clear the continued importance of “local” for music enjoyment and discovery.
Introduction: Local music
The advent of recorded music, and subsequent technological advances in its distribution, have gradually eased the constraints on where and when it is possible to enjoy music. This has fueled the global exchange of music, which in turn has led to the sharing of customs, instruments, and cultures. In the wake of these shifts, a great deal of discussion and research has taken up the question of how this globalization of music has affected countries’ own locally-produced music. However, little work has been done to unpack how exactly listeners and artists define “local” music and how this should affect music recommendation systems.
We seek to fill this research gap through two studies of the concept of “local” in the context of music: a qualitative study that develops a conceptual framework for what listeners and artists consider to be “local”, and a quantitative analysis that studies the effect of incorporating this framework into real-world music recommendations.
What makes music local
To build an understanding of how people around the world currently conceptualized local music, we interviewed music listeners and artists in three cities: Port Harcourt in Rivers, Nigeria; Houston in Texas, US; and Salvador in Bahia, Brazil. We structured these interviews as cultural probes, focused around a series of creative tasks involving the sharing and discussion of artifacts (e.g., playlists, photos, videos, etc.) depicting local music.
The characteristics that define “local” vary over time and place, as people come and go and the musical styles connected with a particular place evolve. Nevertheless, three themes emerged in how participants defined “local” music:
- Music contains signifiers of the place (e.g. language, instruments, beats)
- The artist is from the place
- The artist is currently accessible to people in the place
Signifiers of place take many forms, both lyrical and musical. For example, participants in Houston referenced “chopped and screwed”, a DJing technique developed by area musicians that involves remixing dramatically slowed down samples. In Port Harcourt, listeners emphasized the use of local languages and slang, including phrases like “who goes there”, a subtle reference to the city’s culture.
Naturally, the idea that a local artist must be from a place was expected. Our interviews probed the limits of this idea, exploring both the boundaries of the place itself and the limits of who is considered “from” that place. For example, even an artist who was not born in a place may be considered local after having spent a significant amount of time there. While participants emphasized that the closer an artist was geographically, the more local they feel, the boundaries where local stops ranged from extending up to the country level (Nigeria) in Port Harcourt, the state level (Bahia) in Salvador, and the metropolitan area (Greater Houston) in Houston.
The third theme concerned artists’ accessibility, both physical and emotional, to people in a locale. Participants emphasized the ease or difficulty of attending live performances (e.g. due to cost, frequency of shows), as well as the ease of approaching an artist. To many, this implied that the more popular an artist became, the less accessible and thus less local they tended to feel. However, artists who became popular but maintained a level of emotional accessibility through continued use of local signifiers were still viewed as local.
Artists expressed deep pride and appreciation for their local music communities, which serve as invaluable sources of inspiration, knowledge sharing and mentoring, and of networking and relaying of opportunity. When asked about algorithmic recommendations of local music, many artists emphasized the possibility that, if done responsibly, such recommendations could shine a spotlight on lesser-known artists.
Local artist recommendations
Guided by the findings from our qualitative study, we formulated local music recommendations and delivered them to Spotify users at scale as part of a randomized experiment to study the extent to which local music interests listeners on the whole. We surfaced these recommendations as a horizontal “shelf” of artists on the Spotify Home screen, as seen in the figure below.
To construct this shelf, we first sought to identify local artists and their locales. Our qualitative study described a number of dimensions that contribute to an artist feeling “local”; here we focused upon local artists being described as having geographically distinct listenership. To have “distinctly” localized listening could imply a variety of interpretations. We adopted a straightforward implementation that identifies artists with a surprisingly localized listenership in the statistical sense. That is, if an artist’s Spotify follower count in a place exceeds what would be expected or predicted based on their following in other places, we conclude that this artist is “local” to the place.
To match listeners and artists, we looked at user and artist locations, as well as the user’s musical preferences, which we inferred from their prior behavior. We limited the popularity of artists we considered eligible for recommendation, in order to match interview participants’ description of popular artists as being less likely to be considered local. For each eligible user-artist pair, we calculated a score based on the user’s affinity for the artist’s genres as well as the degree to which the artist is local to that user’s city and region. We then populated a shelf of recommendations with the artists who scored highest for each user.
We ran an A/B test comparing the Spotify Home screen containing this “local artists” shelf to alternative versions of the Home screen containing two control conditions: a shelf of artists chosen using a similar user-artist matching procedure, but which instead drew artists from any locale except the user’s (referred to as “unlocal”); and another shelf similar to the first control, but which ensures that the popularity and genre distribution of the recommendations matches that of the original “local” shelf (“non-local”).
The shelf of local artist recommendations yielded significantly higher rates of user engagement compared to control conditions. The table below shows the relative performance of local recommendations compared to the two baselines, unlocal and non-local, in terms of several measures of engagement: the number of listeners who streamed (listeners sampling at least 30 seconds of a track), clicked (listeners clicking on an artist in the shelf to view the artist’s page), and followed (listeners subscribing to notifications and recommendations pertaining to the artist) one or more of the artists from the shelf. With the exception of just stream counts, local recommendations yielded significantly higher engagement with the recommended artists.
We also investigated whether the outcomes of our experiment varied with listener demographics, by bucketing users into generational categories based on age. Comparing local artist recommendations to the unlocal control condition, we found that the impact of localized recommendations seems to resonate most strongly with younger listeners, as seen in the table below.
In the two weeks following the completion of the experiment, we observed significant sustained interaction between users and artists discovered in the local artists shelf, compared to those surfaced in the control conditions. These results demonstrate the potential for lasting impact of local artist recommendations.
What’s next
We believe our findings demonstrate that local recommendations – and offline context more generally – should form an important part of online content platforms’ strategies. This mechanism for driving discovery contributes to the growth of audiences for lesser-known artists, using literal common ground to connect listeners and creators.
For more information, please see our paper:
Exploring Local Music’s Place in Global Streaming
Samuel Way, Dominika Mazur, Rebecca Kupferman, Megan Walsh, Angelina Tizé and Ben Lacker