Introduction

The Rise and Fall of Music Discovery Algorithms

Music streaming, initially pioneered by Pandora in the US and popularized globally by on-demand apps Spotify and Apple Music, is now the dominant way younger generations listen to and discover new music. With over 100 million tracks[1] available to play on demand—and by offering the implicit potential to discover a wider variety of songs than was ever possible before, the top streaming platforms have captured the ears of over 713 million users worldwide[2].

As streaming customers swelled over the past decade, every record label and independent artist included their works on the top platforms in hopes of finding a larger or more engaged audience. In parallel, revolutionary new recommendation services such as Spotify’s Discover Weekly promised to introduce listeners to previously unheard tracks, personalized per the individual’s listening history. They algorithmically cherry-picked based on similarity to tracks they already liked. In theory, for the first time, musicians could now get direct exposure to potential new fans based solely on their music’s qualitative factors. The promise of better music discovery, and even the initial outputs from the streaming apps, resonated with artists and fans alike, driving music streaming’s growth into a $21.9 billion industry[3].

More recently, however, it would seem the top streaming apps are no longer perceived as the saviors of music discovery and have “lived long enough to become the villain.” In 2023, it was reported that a whopping 74% of all music played on the market-leading streaming platform was concentrated among acts signed to the top four major record labels[4]. More data revealed that millions upon millions of tracks were not being surfaced in algorithms[5]. Given that the significant labels not only license the most popular tracks and artists on Spotify but also hold substantial shares in that company, they exert considerable influence, if not outright pressure, on which artists are promoted and monetized[6].

A decade after gaining initial traction largely due to discovery features, streaming apps have almost completely shifted into a profit-oriented mode, prioritizing podcasts and audiobooks with superior contribution margins to on-demand music[7]. As the top apps work to shift the majority of listening hours toward own-produced podcasts - the modern-day equivalent of talk radio - artists and users have been left with fewer opportunities for genuine music discovery. Moreover, music discovery is now in dire straits with the rise of streaming fraud and a tsunami of white noise and generative AI tracks diluting the overall ocean of real art available on streaming platforms.