Are algorithms the new talent scouts?

As streaming rockets to account for more than 80% of the music industry’s revenue, algorithms are rapidly becoming the new, savvy, talent scouts.

Sending demos to labels is rapidly becoming a thing of the past. Instead, hundreds of thousands of new artists are successfully building global fan bases on their own by uploading their music straight onto streaming platforms.

There are nearly 15,000,000 tracks being uploaded each year, according to London and New York-based A&R scouting platform Instrumental, which focuses on machine learning tools to find high-potential new artists.

With over 40,000 tracks being uploaded to Spotify each day, and around 500 hours of video uploaded to YouTube every minute, Instrumental monitors a staggering amount of music.

Instrumental’s proprietary metrics, including the Instrumental Hot Score, track a range of data points to identify artists with significant potential at the earliest stages of success. Australian-born singer Tones and I was identified as a hot artist just one week after releasing her first single ‘Johnny Run Away’ last year. Having now surpassed a billion Spotify streams worldwide, she still racks up millions of streams per day.

The company analyses around 30,000 influential Spotify playlists and hundreds of the so-called ‘most engaged’ YouTube upload channels, to identify artists who have the biggest potential to become stars on digital services. Instrumental says it finds around 5,000 new artists every week, which it then matches to more than 4,000 ‘micro genres’ to help agents and labels find potential hits. Current clients include Sony Music, image sound and Live Nation.

Other companies are also getting in on the act, including Stockholm and LA-based Snafu Records, which bills itself as the first ‘full-service, AI-enabled record label’. Every week, it uses an algorithm to analyse around 150,000 songs across streaming platforms such as Spotify, YouTube and SoundCloud.

Snafu’s software evaluates tweets, songs added to playlists and blog posts, with the label claiming that it is able to identify ‘undervalued’ artists within days of them releasing their first song, as opposed to the four to six week timeline typically needed by many traditional record labels. The software then ‘scores’ the music to narrow the weekly pool down to no more than 20 artists. Staff at the label then listen to the shortlist to decide whether there are hits in the making.

The ‘sweet spot’ is songs that are 70% to 75% similar to the tracks on Spotify’s Top 200 list, according the label. In short, they are looking for artists who are similar enough to existing ones to gain traction but who have a slight edge to keep it fresh.