New Music Friday Deep Dive: Predicting Commercial Performance of Music By Ariana Grande, Julia Michaels, Lil Pump, Maren Morris, and Rich The Kid

A few weeks ago, we predicted commercial performance of over 1000 new tracks released in Q1 2019. In that post, we summarised how accurately our predictions matched actual performance across all tracks. Today, we’ll dig a little deeper into our results and demonstrate how accurately we can predict not only performance of tracks by different artists, but performance of tracks by the same artist.

To do this, we’ll look at artists who released multiple tracks during Q1 of 2019. Specifically, we’ll focus on five artists representative of our general pattern of results. As we’ll show, our ‘hit rate’, i.e., the proportion of tracks for which our predictions are accurate, is around 70-80%. In other words, given the breadth of tracks in the dataset, we can predict commercial performance of three-quarters of all new music released.

ARIANA GRANDE

Ariana Grande had four songs on Spotify’s New Music Friday playlist in Q1 of this year — 7 rings, break up with your girlfriend, i’m bored, bad idea, and NASA. For each track, we predicted its commercial performance (in terms of Spotify streams) one month after it appeared on the playlist. To make comparisons easier, all predicted and actual performance figures are standardised on a 0-100 scale. Here are our predictions versus actual performance for Ariana:

 
Predicted versus actual performance of  7 rings ,  break up with your girlfriend, i’m bored ,  bad idea , and  NASA , four tracks by Ariana Grande that appeared on New Music Friday in Q1 2019.

Predicted versus actual performance of 7 rings, break up with your girlfriend, i’m bored, bad idea, and NASA, four tracks by Ariana Grande that appeared on New Music Friday in Q1 2019.

 

Our algorithm is specifically designed to predict commercial performance of tracks released and promoted as singles. Interestingly, only 7 rings and break up with your girlfriend, i’m bored were released and promoted as such. For these two tracks, we clearly predicted that 7 rings would be a significantly bigger hit than break up with your girlfriend, i’m bored. That we were able to predict performance of tracks marketed in other ways, i.e., being placed on a popular playlist like New Music Friday, is a good indicator of our ability to deal with more complex promotion strategies.

JULIA MICHAELS

Julia Michaels had three songs on New Music Friday in Q1 2019 — What A Time, Anxiety, and Peer Pressure. Here are our predictions versus actual performance:

 
Predicted versus actual performance of  What A Time ,  Anxiety , and  Peer Pressure , three tracks by Julia Michaels that appeared on New Music Friday in Q1 2019.

Predicted versus actual performance of What A Time, Anxiety, and Peer Pressure, three tracks by Julia Michaels that appeared on New Music Friday in Q1 2019.

 

These results nicely demonstrate how well our algorithm can handle collaborations with other artists. What A Time, for example, features Niall Horan, while Peer Pressure is actually a James Bay track on which Julia Michaels is the featured artist. Even with all these complexities, we were able to accurately predict these tracks’ commercial performance.

LIL PUMP

Lil Pump had three tracks on New Music Friday in Q1 2019 — Racks on Racks, Be Like Me, and Hello. Here’s how our predictions compare to actual performance:

 
Predicted versus actual performance of  Racks on Racks ,  Be Like Me , and  Hello , three tracks by Lil Pump that appeared on New Music Friday in Q1 2019.

Predicted versus actual performance of Racks on Racks, Be Like Me, and Hello, three tracks by Lil Pump that appeared on New Music Friday in Q1 2019.

 

Again, there are some collaborations going on here, with Be Like Me featuring Lil Wayne, and Hello being a track by Ugly God on which Lil Pump is featured. And once more, our algorithm handles the intricacies of these collaborations with aplomb, correctly predicting that Racks on Racks and Be Like Me would not only be significantly bigger hits than Hello, but also hits of similar size.

MAREN MORRIS

It’s not all smooth sailing, though. For Maren Morris, who had four tracks in our dataset, while we correctly predicted the relative success of GIRL, The Bones, and A Song for Everything, we over-estimated how successful Common would be:

 
Predicted versus actual performance of  GIRL ,  Common ,  The Bones , and  A Song for Everything , four tracks by Maren Morries that appeared on New Music Friday in Q1 2019.

Predicted versus actual performance of GIRL, Common, The Bones, and A Song for Everything, four tracks by Maren Morries that appeared on New Music Friday in Q1 2019.

 

Common, however, was a promotional single, not a full single, while A Song For Everything was never released as a single at all. Of the two tracks that were full singles — GIRL and The Bones — we correctly predicted the significantly greater success of the latter. Thus, while our algorithm is quite robust, and can overcome some of the complexities associated with different types of release, it is not infallible.

RICH THE KID

Finally, Rich The Kid had four tracks on New Music Friday in Q1 of this year — Facts, 4 Phones, Tic Toc, and Fall Threw. Here’s how our predictions compare to actual performance:

 
Predicted versus actual performance of  Facts ,  4 Phones ,  Tic Toc , and  Fall Threw , four tracks by Rich The Kid that appeared on New Music Friday in Q1 2019.

Predicted versus actual performance of Facts, 4 Phones, Tic Toc, and Fall Threw, four tracks by Rich The Kid that appeared on New Music Friday in Q1 2019.

 

Despite Facts being a four-way collaboration between Chantel Jeffries, YG, Rich the Kid, and BIA, and Fall Threw never being released as a single, we were able to fairly accurately predict the relative performance of these four tracks.

THREE QUARTERS OF ALL NEW MUSIC

These selected artists and tracks are broadly representative of our results, giving us confidence that we’re able to accurately predict commercial performance of around three-quarters of all new music released. Our algorithm is not perfect, but it is significantly better than chance. And we’re constantly developing it to improve the accuracy of our predictions.

If you have any questions or comments about these results, or would like to learn more about how we calculated them, we'd love to hear from you! For more general info about how Hyperlive quantifies listener engagement and forecasts future performance, check out our Home or About pages.

About Hyperlive. Music isn’t something we just listen to — music is something we experience with our heart and soul. Hyperlive captures this experience by modelling a range of neurobiobehavioural responses to music as well as the psychological processes that underpin them. This gives us a deep understanding of what drives musical engagement on a fundamental level, allowing us to quantify, model and predict that engagement — and the musical features that motivate it — with unmatched levels of precision.

Hyperlive