top of page
Search
Writer's pictureAndHeHadAName

Discover Strongly

Updated: Aug 24, 2021

Behind every great playlist maker is a great algorithm.

Last month in July I went to go see a singer called Bassh. As I noted in my review, it appeared I was the only person who was there that was not either directly connected the artist, or brought by someone connected to the artist. So what caused me, a single individual to find themselves at a basement performance to a friends and family show?


The answer: Spotify's Discover Weekly (DW) algorithm. Approximately 3 years ago in August of 2018, a song called Body appeared on my DW. A beautifully produced track with a sexual edge, it then sat there on my Only the Lonely Psychedelic Glam Rock list, collecting a few hundred followers, but otherwise existing in obscurity. Yet, if my DW had never presented me with this song, or if I had just ignored it, I never would have placed Bassh into my BandsInTown, and never ended up in that show.


There are a lot of bands in the world, and I mean a lot. Since you started listening to this a dozen songs have dropped on BandCamp and there are half a dozen conversations occurring between friends that will result in the formation of an indie rock group. Each day hundreds of compositions are released in thousands of different sub genres from artists of every age and every location (even places that aren't NYC, LA, or Nashville). From established bands and new bands. EPs, LPs, Demos, Covers. City Pop, Midwestern Emo, Mood Rock, Dude Rock, 90s Revival, Soul Music, Ted Nugent (jk), R&B throwback, much better music in the style of Massive Attack. How is a normal human (or even lesser musical gods) supposed to keep up? There is no possible way for an individual to go through album reviews, other playlists, listen to new and old releases, or rely on good ol fashion word of mouth for this to be efficient.


To sort through this relentless jungle that is new music, the only way to give yourself even a chance is for the music you are most likely to like be presented in front of you. Traditionally done through the radio function which has been a part of digital music streaming since its inception. Who can forget creating their Beatles/Stones station and then their Wu-Tang/Nappy Roots station on Pandora? But that had its limits. The recommendations remain too generic as it settles on popular consensus to select the tracks. But you aren't a consensus, you are a you.


You aren't a consensus, you are a you.

This is where Discover Weekly truly differentiates itself from the only ways. Now while the exact specifications of the algorithm are not fully known, in basic terms what it does is takes advantage other Spotify user's preference for song similarities (e.g. if a song appears on one playlist with another, Spotify will assume the songs are related), but takes it one step further by also determining which specific Spotify users are most likely to make musical associations you agree with (for instance if you ever went through a mashup phase, you get auto rejected by my algorithm). By allowing you the ability to select not only which songs you personally like and which associations you agree with, the algorithm becomes individualized even if the process is generalized in nature.




This is me with my Discover Weekly Algorithm, finding new indie tunes

In this way DW almost serves as a musical extension of your brain, and can be though of a form of cybernetics (turning part of your physical or mental self into a machine). Now if this concept seems almost sci-fi, it is because it is. The ability for an machine learning algorithm to reach its tendrils across both time and space while relying on the collective ability to its million of users to subset people into their own consensus group is nothing short of alien . And I could go on (and will in later articles) about how powerful this algorithm truly is. For now, it is enough to understand that not using Spotify's Discover Weekly algorithm is self-sabotage.


184 views0 comments

Comments


Post: Blog2_Post
bottom of page