For Snafu Records, AI is the Answer to Find New Artists - Rolling Stone
Home RS Pro Music Biz Features

This New Record Label Is Using an Algorithm to Find Its Artists

Newly launched label Snafu is doubling down on a tech-assisted talent search

snafu records algorithmssnafu records algorithms

Snafu Records

Most aspiring artists know there’s a one-in-a-million chance of success, but if the newly launched Snafu Records has its way, it’ll be one in 150,000. That’s the lofty number of songs the label sifts through each week to find their new artists. Their talent scouts, though, aren’t doing it alone; they’ve got the help of an AI-powered algorithm.

Snafu is banking on what it calls AI music discovery. Rather than relying solely on a team of A&R scouts to scour the scene for the next big thing, the label is looking to some code to make its job easier.

Snafu’s algorithm narrows the 150,000 undiscovered weekly songs to 15 to 20 of the best options for the label’s four-person A&R team to decide from. The algorithm analyzes factors like a song’s growth volume, fan engagement — such as positive YouTube comments — and how similar a song may be to another hit track. (The song should be 70-to-75 percent similar to the average song on Spotify’s Top 200 at a given time; different enough to be unique, but similar enough to follow the “rules of a hit,” as Snafu CEO Ankit Desai says.)

Speed is a large part of the label’s goal with its AI approach. The team is typically notified within two to nine days of a given song’s release, significantly quicker than Snafu’s four-to-six week estimate it takes for a song to get on a label’s radar. As songs can often go viral outside of a label’s control, much of the process becomes outbidding labels for whichever artist exploded on the scene seemingly overnight. Snafu says getting to potentially viral songs early — the goal is to get them before they hit TikTok fame status — and hearing songs first makes it much easier to predict and mine a hit.

Snafu isn’t the only label using analytics and data to find artists. Warner Music Group purchased a similar algorithmic A&R tool called Sodatone in 2018, and almost any A&R scout has already been using different analytics from Shazam and Spotify to predict what’s on the come-up. But where the majors use tech as one tool in their arsenal, AI is deeply embedded into Snafu’s DNA as a company, its executives say.

“It felt sometimes like we were taking 21st-century technology and then plugging it into machinery that was really built for success in the 20th century,” Desai, who previously ran data analytics at Universal Music Group and Capitol Records, said of bringing data analysis to labels. “They’ve definitely realized that this is the future, and they’re also moving towards that, but I think like all of them, they’re facing problems that any big company faces during times of change, which is that they’ll they’ll take a bit longer to get there than Snafu that’s a native in this space.”

The company, in development since 2018, has signed 16 artists. While it has seen quick growth from recent signings like Costa Rican electro-pop singer-songwriter Mishcatt, whose Spotify numbers have grown by tens of thousands in a matter of months, Snafu still needs to make hits to prove such a data-heavy approach can actually work.

“It felt sometimes like we were taking 21st-century technology, and then plugging it into machinery that was really built for success in the 20th century.”
—Snafu CEO Ankit Desai

In an industry as historically averse to change as the music biz, the reliance of more tech may be met with skeptics uncomfortable with the idea of a bot deciding who and what fans are going to listen to. But Desai, who heard skepticism from some industry members before launch, is wary of such a critique. Some critics “looked at A&R as this black box you couldn’t touch and [that] applying data to it in some sense could take away the magic,” he says. “But we look at it the other way. Who are we to be the gatekeepers of what people should listen to? At the end of the day, what the numbers and the data are showing are really what the world is consuming.”

Several A&R reps contacted by Rolling Stone held a generally accepting view of adding technology into the discovery process as long as it remains a helpful tool in finding artists rather than the primary factor. Songs like “Old Town Road” changed everything and have pushed labels to think more about the viral approach, RCA A&R director Karl Fricker says, with many labels looking to reverse-engineer the “Old Town Road” process to find hints of how to strike gold again.

“It’s a game changer; the idea of someone coming literally from obscurity to the biggest No. 1 song,” Fricker says of Lil Nas X. “But there’s no algorithm, no forecasting that could have been done to forecast those events.” Still, labels will continue to lean more in the direction of experimental startups like Snafu. “I don’t see why it wouldn’t,” he says. “This data is a reflection of what’s in the marketplace. It’s an empowering thing to know this, and what it does is reinforces your gut. I can’t see a reason you wouldn’t acknowledge the general public. How could you understand what a teenager is feeling and dictate what they should like? That’s what this tech provides.”

“This data is a reflection of what’s in the marketplace.”
— RCA A&R director Karl Fricker

Ryan Faus, vice president of A&R research at Arista Records, said Snafu’s concept is a good addition to the industry given the sheer amount of content that scouts pour through to find new material. “The amount of music that we have to be aware of in our quest to find superstars and hits is staggering, and so I think technology is going to be one of our best tools to filter down that pool,” he says. “At the end of the day, there’s been a lot of nervousness in the public, and perhaps in the industry, about the reliance of computers or AI in this space; and to me, the perfect alchemy is a blend of using the data. But I don’t think there’s been a replacement for ears of an A&R to understand the potential of a song or artist.”

Faus said that four to five years ago, such an embracive perspective toward tech may have been criticized, but today, labels understand the need to find new tools to beat competitors and cut through the vast amounts of new music. “The industry has evolved rapidly in the last few years in terms of our ability to read, interpret, and work with information, especially when it comes to identifying talent, but I think we’re all looking for room to grow in finding what our next steps could be in terms of getting better at this process,” Faus says. “There’s probably a bit of an arms race between the labels and even the distributors developing tools and methods and ways we can kind of digest all the things we can know about what’s happening out there. I don’t think that’s going to stop or slow down.”

“Years ago, it would take us months to try and close a deal,” adds Yaasiel “Success” Davis, VP of A&R for Atlantic Records. “Now we’re able to get things done quickly because there’s so much consumption. That’s why these technology companies started becoming a big asset.”

While Snafu plans on using AI and data analytics more than a typical label, Carl Falk, Snafu’s creative director and a songwriter and producer with credits on Nicki Minaj and One Direction hits, was quick to dismiss the idea that it’d let its algorithm ever fully replace human talent scouts.

“It would be a big mistake if we only looked at numbers,” Falk says, likening the algorithm to an extra set of eyes for its A&R team. “It’s important to look at the algorithm as a helping tool to find music rather than to fix or enhance it. If we don’t have the ears, than the eyes will suffer from it. I think we’ve found the magic combination of using the algorithm in such a way that we’re not dependent on it to say, ‘This is a hit.’ But it gives us a really quick customer survey on what to bet on, and that’s really important today.”


Powered by
Arrow Created with Sketch. Calendar Created with Sketch. Path Created with Sketch. Shape Created with Sketch. Plus Created with Sketch. minus Created with Sketch.