Spotify just patented it's own A&R technology to predict breaking artists. Should labels be concerned?

A new track is uploaded to Spotify nearly every second of each day. So using the service as a talent scouting tool can sometimes feel like finding a needle in a haystack. Spotify knows this – and it’s working on something to remedy it. MBW has discovered that Spotify has invented technology that predicts which “relatively unknown” artists on its platform “are likely to break” in the near future. According to a new US patent granted on Tuesday (March 1) and obtained by MBW, Spotify has developed what it calls a “system and method for breaking artist prediction in a media content environment”. Spotify’s new invention works by determining “one or more early adopters” from its user base and then collecting data from their listening patterns. The system can then predict which artists are more likely to break based on the listening patterns of these early adopters. The patent explains: “A user who requested playback of media content from a plurality of breaking artists is determined and assigned the role of early adopter. “The breaking artist prediction logic can thereafter predict future breaking artists based on further playback requests from an early adopter interacting with their media device via the software application. Spotify notes that a user flagged as an early adopter “need not have any knowledge that they are an early adopter, as the media server responds by streaming the requested media content, as it responds to all users”. It’s A&R, folks – but the listeners are doing all the work. Even if they “need not have any knowledge” they are doing so. Within the filing – originally filed in 2016 and which you can read in full here – Spotify uses Figure. 4 (pictured above) to illustrate its system for predicting breaking artists. Spotify says that the system includes “a back-end media server system” and that this media server “can include a breaking artist prediction logic”. In essence, this “prediction logic” watches consumers that Spotify has identified as “early adopters” to see which breaking artists they are playing and re-playing. Spotify explains in the filing that, “the breaking artist prediction logic can determine artists that have increased in popularity in the past based on […] historical data”. Spotify explains that an artist can be deemed to be “breaking” if “the popularity of the artist exceeds a minimum playback threshold while having experienced growth in popularity that exceeds a minimum growth threshold”. Spotify suggests as an arbitrary example that “a minimum playback threshold” could be 200,000 playback requests across the platform in one month. A “minimum growth threshold”, meanwhile, could be a 300% month-to-month increase in playback requests. Spotify underscores that these metrics are “merely exemplary, and a minimum playback threshold may be higher or lower”. Additionally, according to the filing, numbers for determining whether an artist can be deemed breaking or not can vary “between genres, formats, and audience demographics”. “For example, a minimum playback threshold for a breaking pop singer may be 200,000 playback requests in a month,” explains Spotify, “while a minimum playback threshold for a breaking pianist that records classical pieces may be 20,000 playback requests in a month”. Spotify argues in its patent that “if a service provider can predict that an artist will break out”, then that service provider could also potentially “seek to partner with the artist early for possible promotional and recording deals”. (Any major record labels reading: yes, that does indeed indicate that Spotify is exploring the prospect of using this tech to ink early-stage “recording deals” with about-to-break talent. We expect this is all totally fine to you.) A partnership between a “service provider” like Spotify and an artist, adds the company’s patent, could “also provide an opportunity for the service provider to push media content by artists to playlists and radio stations to grow the artist’s fan base”. By doing this, says Spotify, it would play a role in “associating the service provider with ‘hip’ or ‘fresh artists'” and therefore “establishing the service provider as the source for discovering new and interesting media content”. Using streaming technology to predict breaking artists in the music business is nothing new, of course. London-born Instrumental, for example, is well-known for its AI-driven A&R scouting platform, which was able to spot huge hits by the likes of Lil Nas X and Tones & I months before those artists signed major label deals that helped their careers take off. Instrumental, in which Warner Music Group acquired a stake in 2015, uses proprietary machine learning combined with human intelligence to comb streaming services and discover around 5,000 new artists a week. Another talent scouting tech platform, Travis Rosenblatt’s Meddling, says it has been used by the likes of Republic Records and Interscope to discover emerging artists. And AI-driven platform Musiio runs a proprietary ‘Hit Potential Algorithm’ which it says can automatically classify and categorize new music – as well as measure its hit potential – from the content of the music itself. Spotify seeking to develop what is essentially its own A&R tool to identify breaking artists, and then – as per the patent filing – potentially “partner with the artist early for possible promotional and recording deals” adds an interesting new layer to the power dynamic between Spotify, artists and labels. A patent for a native Spotify talent scouting tool is a potentially game-changing prospect, especially when viewed in the context of rival SoundCloud recently launching a new top-tier of artist service offerings – referred to as its “roster” – which will see particularly successful independent artists ‘sign’ direct to SoundCloud. Remember, it was only in 2018 that Spotify started offering artists direct distribution deals – including paying out advances, but then backed out of the plan. Spotify also launched its own DIY distribution service in 2018, only to abruptly pull the service just a few months later. Could the technology detailed in this patent signal a potential change in Spotify’s approach in future?