Some of us are avid music listeners, constantly spending the day listening to playlists and mixes whether it’s for relaxation, for jamming, or as background noise to help us get through a particularly challenging assignment. Some of us also love sharing music with our friends. Inevitably, one might create a playlist containing their current favourites in the hopes that the receiver of the playlist might add a few of those songs into their own. Before sharing them, we’ll commonly preface them with “My playlist might be a bit all over the place. It’s got a little bit of everything so you might not like all of it”. I wanted to know; are the artists I’m listening to really that different from each other? Can we identify communities within the artists I’m listening to? To help me answer this question, I’m using this tool called the “Spotify Artist Network” to help me visualize the network of artists. For this blog, I’m going to use American alt-rock band Wallows as the root of my network. This is the resulting network.
It is important to note that the clusters represent related artists and not artists who have worked with each other. How does Spotify determine related artists?
The tool does a great job of organizing artists into clusters for us. The root node is in black and the node size and color indicates popularity of the artist. In the center cluster, we have related artists like Conan Gray, girl in red, Beach Bunny, Cavetown, mxmtoon. The sounds of Wallows’ songs are backed with the sounds of snares and crash cymbals and arguably can not be compared to the soft ukulele instrumentals of mxmtoon’s bedroom pop. Despite the difference in sound, Spotify’s algorithm has decided that they are closely related. This may be due to the fact that their songs are commonly used in TikToks, and they’re also popular social media personalities. In a distantly related cluster, we have hard rock artists who are commonly compared like My Chemical Romance, Mindless Self Indulgence, and Pierce the Veil. In another cluster, we have intimate indie-pop artists like Peachy!, Shiloh Dynasty, and khai dreams. All of whom are commonly featured in hours-long lo-fi mixes on YouTube.
Spotify does have an actual answer to the question. In short, Spotify analyzes the number of fans shared by two artists proportional to their total number of fans and also the way they are described on other media (i.e. blogs, magazines, social media). The number of shared fans has a higher weighting but artists getting a surge of attention in the media are also a significant factor. This is clearly shown from the observations above. Some of the clusters are highly connected due to the genre and the common comparisons in media between them. Other circumstances like social media trends and challenges are also heavily influencing how the artists in this network are clustered.
So yes, we can definitely find communities and see how these seemingly disconnected artists share a large portion of the same fans. In addition, using these clusters, we can determine which songs could be classified as a different genre, but we would still enjoy it. Of course, there are also clusters consisting of a single node between clusters so those songs can be seen as an entry to another cluster.
For further technical investigations of Spotify songs and artists, one can look at the Spotify Developer API.
For more information on the “related artists” algorithm, check out Spotify’s blog post: How “Fans Also Like” Works.