Previous Submission

Spotify uses listener data to predict, test, and create hit songs.

Next Submission

Spotify uses listener data to predict, test, and create hit songs. "By the time a song lands on Todays Top Hits or other equally popular sets, Spotify has so relentlessly tested it that it almost cant fail.

With a recent IPO, Spotify has captured immense value as a music streaming platform. According to the company, the platform has 157 million active users, over 35 million songs, and over 2 billion playlists (1). Recently, there have been numerous articles published about how the company utilizes ‘big data’ to create more customized playlists for users and discovery. The company looks at songs users listen to, data about song characteristics, and many other metrics in its endless pursuit for more custom catered music. However, less frequently discussed is Spotify’s use of data to spot ‘hits in the making’ and their ability to ‘create a hit single’ through their platform. The company’s ability to ‘create a hit single’ through data has also shifted its negotiating leverage with record labels.

Nick Holmsten, the service’s head of shows and editorial, claims he could dig into the data and tell you which new song will be a hit in six months.” According to multiple sources, Spotify treats every song on the service “as a beta test”. The service monitors the number of plays songs receive and utilizes algorithms to spot songs that are starting to gain popularity. Once a song has reached the algorithms threshold, the song is funneled into a series of tests to discover its ‘hit potential’. “Think of it as the moneyball of music, a ruthlessly data-driven approach to introducing listeners to songs.” Spotify uses a strategy it calls “playlisting” (2). Through this process, songs are added to Spotify created playlists with different numbers of active followers. Spotify considers playlists to be a fundamental competitive differentiator of its product. “Listeners now spend about half their time on Spotify listening to playlists, either of their own creation or curated by Spotify’s editors and other tastemakers.” (2)  The company posts ‘songs with hit potential’ onto playlists with some followers and if the “numbers look good”, the song is upgraded to playlists with more subscribed listeners and enters a new round of competition with other top performing songs. This process continues until the most successful songs hit the major playlists. “By the time a song lands on Today’s Top Hits or other equally popular sets, Spotify has so relentlessly tested it that it almost can’t fail.” (2)

In an article titled ‘How Spotify Creates Hits’, Rolling Stone Magazine recounts the story of Electro-Pop singer Lauv and his smash single ‘The Other’. The artists discusses how he tracked Spotify adding the songs to bigger playlists and how it radically increased streams from 8 to 100 Million. According to Rolling Stone, this increase in virality led to the artists first headlining tour.

Interestingly, record labels have caught wind of this new tactic. This data play has shifted the company’s negotiating leverage with record labels. Specifically, the ability to pick and choose which songs get heard through these playlists is already being used as a negotiating tactic by the company to obtain better terms from labels. “Record companies that offer it lower royalty rates, don’t do exclusives with its competitors like Apple Music, or get their artists to release special re-recorded Spotify Sessions of their hits could see their artists placed more prominently in Spotify’s playlists and their audiences grow. Labels that don’t play ball with Spotify might sublty notice they’re not getting the same playlist love.” (3) 

 

(1) https://press.spotify.com/us/about/

(2) https://www.wired.com/2017/05/secret-hit-making-power-spotify-playlist/

(3) https://techcrunch.com/2017/03/18/dictate-top-40/

7 thoughts on “Spotify uses listener data to predict, test, and create hit songs.

  1. Fascinating read, Jesse. It’s scary how good Spotify’s algorithm knows us by this point. It also gives them access to an interesting set of opportunities: they can choose to become their own supplier by signing artists directly, turning into a label themselves and finally improving their profitably. For sure they have enough data to craft very successful hits. However, their value proposition as an aggregator would be eroded if labels see them as a competitor and they pull their artists out of the platform. I wonder who has more to lose, and how future contracts between Spotify and the Big Three will look like in a not so distant future.

  2. Hi Jesse, thanks for the great post! By leveraging the data, Spotify knows customers well and make them addicted to the platform. I agree with you that the use of data to predict and even create hit songs gives it a great leverage in handling the power dynamics with music labels. This ability sets Spotify apart from its competitors like Pandora/Apple Music and helps the company negotiate a lower royalty cost with labels. I just realized that Spotify also provides an app “Spotify for Artists”, allowing artists to get access to listener data, like who are they, where are they, and which songs are most popular and help win new fans… The insights from data analysis can help empower independent artists to create more hit songs and thus benefit listeners and the company.

  3. Thanks for such an interesting read, Jesse! It’s amazing how Spotify uses data to its advantage in providing personalized services and increasing its power in negotiation. I wonder if they have the potential to gradually move up the value chain – from music-streaming to the music-creation, like what Netflix did with House of Cards, if the argument is when you have enormous amount of data and the channel as well as customer base to relentlessly test, it’s really hard to miss. However, the example of Amazon says otherwise – they completely replies on user preference data and created a show called Alpha House, which fell short and landed at 7.5 on the curving (vs 9.1 rating for House of Cards), marking it as a completely average show. I guess there’s more art and nuances into it when it comes to creating content than predicting the next hit. It would be interesting to see if Spotify will make the move.

  4. Really interesting, Jesse. Fascinating how something like this really has the potential to disrupt the power dynamics in the industry. It had me thinking as I was reading your post… you can almost imagine a situation where Spotify could know so much about what types of music are trending at a particular moment that they could automatically produce a song/beat that would appeal to listeners (i.e., using particular pitches, tempos, etc.), especially in more electronic genres. I hope that all this doesn’t stifle creativity and impact the diversity of the music that’s released or promoted through the app!

  5. Thank you very much for sharing Jesse. I wonder though how much Spotify could keep using this as a leverage to the music label. The way I see it is that each song has its own intrinsic value that would go viral if it gets the chance. If Spotify keeps customers away from good songs too much due to the lack of good terms then over time it will destroy its own value. I think it will be quite scary for Spotify if multiple of good music labels partner up to negotiate with them.

  6. Thanks for the interesting post! I mentioned the exactly the same idea in my comment of another Spotify posts. Then when I sit down and reflect after reading you post, I start to question myself whether this is a dilution of art itself. If we’re only talking about making music, nowadays the AI deep learning is already able to create music by itself. But I still think that music is not just a combination of words, notes and rhythms.. It is all about human emotions. How can you produce human emotions through data and machines?

  7. Fascinating the amount of market power that Spotify now commands! I’m a huge devotee, and a frequent playlist creator, and I do have to admit that the song suggestions at the bottom of playlists has made me a bit lazy—now I can just create about half the playlist then listen to their recommendations to fill out a good set. It’s also been interesting to see how artists and labels promote their music. I have one jazz playlist with over 6,000 followers, and I’ve gotten several messages from artists and their labels about adding their songs—clearly they see value in using the power of playlists to get on people’s radar.

Leave a comment