Skip to main content



Music Taste as a Market

Now we all know the already established markets that revolve around music such as record labels, music streaming devices, music award shows, etc. However, here we will examine people’s music taste on how that is in of itself its own market. First, we must define what we mean by “Music Taste”. It is perhaps more appropriate to label it one’s music choice that is: the songs a person decides to listen to more than once and by their own accord. One way to think about it is the music one might add to their collection whether that be on Spotify, Youtube, or any type of physical or virtual collection.

Now we will mainly try to focus on the music listener and thus this will closely resemble a one sided market where the music listener has certain preferences over what they desire to listen to. The preference list a person will have will depend on how much they like the song given their particular taste. Here it can be very interesting to think about what an agent’s preference might be dependent of since usually the music we like is strongly correlated to what we listened to growing up or perhaps what people we care about listen to. We can also think of matching a song to an agent as bringing the agent some sort of positive utility. Whereas, it will follow that an agent will prefer a song to another if it will bring them more utility than another song. One interesting point to examine is that musical taste is often dynamic and many times the utility of a song can change given if that person has been “matched” (has listened) to other songs are similar. For example, a person who doesn’t really like classical music will have low utilities on classical songs. However, say they take a class that studies classical music and through listening to different pieces and learning about it they find a newfound liking to the genre thus, increasing the utility of classical songs.

Many music streaming services are actually very dependent on how well they navigate this market. Pandora, Spotify, and Soundcloud all have features such as “suggested songs” or “song radios” that try to match agents to songs they might like. Here these services take a different approach on how they handle this dilemma. The approach in many ways can be compared to the model that the service decided to implement. Spotify’s Discover Weekly approach to this market is agent driven. It utilizes information of other agents (other people’s playlists) and information about your taste as well as machine learning to try and recommended new songs for you that you will like. On the other hand, Pandora took a more musically driven approach. Pandora’s approach relies on something called a Music Genome. “Pandora relies on a Music Genome that consists of 400 musical attributes covering the qualities of melody, harmony, rhythm, form, composition and lyrics.” It is important to notice that Pandora has limited Music Genome to contemporary artists and has still yet to approach a model for classical composition. That is they are trying to model musical composition versus people’s musical taste composition.

Which model is more effective? Try both and see which you prefer.

Sources:

https://qz.com/571007/the-magic-that-makes-spotifys-discover-weekly-playlists-so-damn-good/

https://computer.howstuffworks.com/internet/basics/pandora.htm

Comments

Leave a Reply

You must be logged in to post a comment.