🟪⬛ 11: Learn how the Spotify AI categorises your music in 30 minutes
Get a basic understanding of how the algorithms work.
Entire industries are built around the world’s best kept secrets: Algorithms. Spotify comes to mind as a tool where algorithms can have a huge effect on a musician’s career. We can’t tell you how the algorithms works, simply because we can only speculate. What we can do is help you understand how the algorithms obtain information about your music, using the official Spotify API.
Today: Learn how the Spotify AI categorises your music in 30 minutes
Follow the exercises below and dive into the Spotify Developer Console with your artist profile and most popular music.
Write down your observations. Does it match your artist brand?
Of course, Spotify works with highly skilled humans of flesh and blood, who are responsible for editorial playlists with millions of followers and marketing campaigns with a dazzling reach. But more than that, Spotify relies heavily on artificial intelligence (AI) to serve the tastes of every niche.
Some insights in how Spotify uses AI and machine learning:
Spotify’s most popular playlists are composed by algorithms: Discover Weekly, Release Radar and your Daily Mix playlists.
Spotify’s homepage is personalised through an algorithm and recommends new music to listen to.
By default, Spotify never stops playing unless you press pause: The Spotify Radio algorithm creates a collection of songs based on the music you’re playing.
Spotify’s Fans Also Like feature, a.k.a. related artists, is composed through an algorithm.
A basic understanding of how Spotify categorises your music is valuable knowledge, and can help you come up with strategies to influence the outcomes of Spotify’s algorithms. Let’s dive in!
Spotify’s publicly available Developer API Console provides a lot of behind the scenes information about your music. Don’t worry, we’ll save you the developer-lingo and try to explain the steps below as simply as possible. Jump into the Artist Lockdown Challenge community on Discord if you need help.
Before we start:
Make sure you’re logged in to your Spotify account in your web browser.
You will need a so-called access token the first time you run this. It's easy: Press the big green ‘Get Token’ button on any of the web pages below. A popup appears. Don’t tick any of the checkboxes, just click ‘Request token’. That’s it!
Artist
Let’s start with some insights into your artist profile. Below’s manual shows you which genre(s) Spotify assigns to your artist profile, the popularity of your artist profile and the artists who you relate to according to Spotify.
First, obtain your Spotify Artist ID. Go to your artist profile in the Spotify Desktop app > click the three-dotted button next to the ‘Follow’ button > Share > Copy Spotify URI
Paste the result somewhere in a text document. You’ll see something like
spotify:artist:0LcJLqbBmaGUft1e9Mm8HV. The last part is your Artist ID.Open the Get an Artist page of Spotify’s Developer Console in your web-browser.
If the OAuth token box is empty, click Get Token and follow the steps described in Before we start.
Next, paste your Spotify Artist ID in the box labeled “id*” > Click ‘Try it’.
Look at the results in the black column on the right. Inside the code you’ll see genres and popularity.
In Spotify’s documentation you can find more info about the results.
Track
Let’s move on to the music itself.
Obtain the track id of your most popular track. Go to your artist profile in Spotify’s Desktop app > click the three-dotted button next to the track name of your most popular track > Share > Copy Spotify URI.
Paste the result somewhere in a text document. You’ll see something like
spotify:track:0GjEhVFGZW8afUYGChu3Rr. The last part is the track ID code. Make sure to obtain a track ID, and not an album ID.Open the Get Audio Features for a Track page in your web-browser. Paste your Track ID and click ‘Try it’.
The so-called audio features of your track pop up. Examine how Spotify indexes your track on mood-specific variables like the danceability, liveness and valence of your music (with 0 typically being low and 1 being high). Spotify’s documentation goes into more detail.
Now, paste your Track Id on the Get Audio Analysis for a Tracks page and click ‘Try it’.
Now it gets really interesting. The audio analysis shows you how the AI “listened” to your music and translated audio to data. It shows you how Spotify’s AI chopped your track into tiny pieces of a few (mili)seconds, and analysed that piece of music to variables like loudness, tempo, key, confidence, timbre and so on.
Take a moment to reflect on the data. Understand how Spotify categorises your track and understand the decisions that are being made by machines. How does what you've seen compare to your artist brand? Does it match? Write down your observations.
Want more? The left column of the Spotify API Console shows you all available end points. Take a look at the playlist end point for example, and learn how Spotify judges the playlist you made yesterday.
Today’s learnings are not the only way Spotify feeds their algorithms. This article, where Spotify explains how the “Fans Also Like” feature works, provides some insight into how Spotify analyses web pages about music to obtain data about music.
We know, this is geeky stuff. 🤓 But hey, this challenge is all about digital strategy, right? We think it’s important to have some basic knowledge of how this works. It’s not just about knowing the right gatekeepers anymore, machines might be already the most important influencers in music discovery.
Make sure to share your observations with the Artist Lockdown Challenge community on Discord. We create dedicated channels for each daily task.
🟪 Need help? Join us on Discord and get help from the Artist Lockdown Challenge community.
⬛️ Task done? We’re happy to share the results: tag @artistlockdownchallenge on Instagram and we’ll repost.