Predicting Performance on Spotify — South By South West + BBC Music Special

 
 We were able to predict streaming performance on Spotify for a range of British artists playing SXSW 2018 with around 80% accuracy by analysing the musical content of their songs.

We were able to predict streaming performance on Spotify for a range of British artists playing SXSW 2018 with around 80% accuracy by analysing the musical content of their songs.

 

PREDICTING ARTISTS' PERFORMANCE...

It's all well and good being able to forecast performance of songs by international superstars, but what about lesser-known or up and coming artists? With South By South West in full swing, and a compliment of artists performing at the British Embassy courtesy of BBC Music, we thought we'd let our algorithm loose on some emerging British talent this time around.

The BBC Music event kicked off last night, and continues until Saturday, so we began with the first night performers — Himalayas, Jerry Williams, Pale Waves, Rachel K Collier, and Touts. We analysed every single released to date by each of these artists, then applied our music-focused algorithm to predict each track's performance. And just for a change, we decided to predict Spotify streams only on this occasion, just to see how good our algorithm can be when tasked with predicting platform-specific performance.

...ON SPOTIFY...

With no prior knowledge of actual performance, we used our algorithm to classify each track into one of five performance ranges: 0-500k streams, 500k-1M streams, 1-5M streams, 5-10M streams, and 10M+ streams. We felt these ranges offered a good combination of granularity and relevance for these artists. Having predicted each track's performance, we then compared our predictions with today's Spotify play count. So how did we do? 

Of the 15 singles released by these artists, our predictions matched actual performance for 14 of them. In other words, we were able to predict performance on Spotify with over 90% accuracy.

Encouraged by this, we turned our attention to tonight's performers – Frank Turner, Gaz Coombes, Jade Bird, Nina Nesbitt, Rhys Lewis, and Sam Fender. Again, we used our algorithm to classify every single released by these artists into one of the same five performance levels. Having predicted performance, we then compared our predictions with today's Spotify play count for each track.

Of the 53 singles released by these artists, our predictions matched actual performance for 32 of them. In other words, we were able to predict performance on Spotify with over 60% accuracy. Combined with our first night predictions, that gives us an overall average of 77% accuracy.

...WITH UNPARALLELED LEVELS OF ACCURACY

To put it another way, 8 times out of 10, our algorithm correctly predicted performance on Spotify, the world's biggest music streaming platform, for a range of British artists at different stages of their careers. And remember that our algorithm is completely focused on the music. We don't know the extent to which each track was promoted, we've not examined potential effects of music video or possible influence of featured artists, and we don't utilise co-varying factors such as social media activity when making our predictions.

As we've noted previously, considering the many factors one might expect to influence how successful a track becomes, that so much of that success can be traced back to the music itself is quite remarkable and, from an artists point of view, presumably rather encouraging.

WANT TO KNOW MORE?

To learn more about how Hyperlive quantifies listener engagement and forecasts future performance, check out our Home or About pages, or get in touch!

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.