How deep can we dive into music with data science?
Let's take a look at Hollywood. Instead of guessing a movie's box office revenue, the world's blockbuster powerhouse figured AI would do a better job.
ScriptBook — a tool that approximates box office revenue through screenplay analysis — uses machine learning to process unstructured data. In other words, the startup analyses the movie's screenplay and its visual story. ScriptBook's predicted The Passenger to earn $118 million. How much did it actually make? Around $118 million.

As the film industry and the world of music are tightly intertwined, can we do the same for music?

Back in April, I tried doing a data science project which would predict a song's place in the charts, but it was very, very subjective. As far as my research went, I didn't find a single startup that was doing at least something along those lines — only an article on Medium, where the writer was predicting whether the song is hit or not.

Of course, based on those results, you could apply this algorithm to songs that are just being released. But while a №1 song is more likely to bring more revenue that a song placed below the top 100, the song's chart position does NOT equal high revenue.

What is more interesting is that the music streaming juggernauts — Spotify, Pandora, Apple Music, SoundCloud — are not doing anything about this.

Spotify could potentially predict the revenue of their artists' songs, and pinpoint their weak areas. At the same time, SoundCloud can create a new feature where they help emerging young musicians get more recognition by implementing AI algorithms.
Made on