During the season of Spotify Wrapped, If you’re anything like me, you might have been surprised to check your wrapped and learn that there are hundreds of genres that you supposedly explored? As it turns out, Spotify now has 5782 (sub)genres in its collection, so some of them are a little odd or curiously precise. Even when we’re usually familiar with broad labels such as “pop,” “indie,” and “metal,” there are a number of super-specific titles that may astonish even the most educated listeners. After viewing my friends’ Spotify Wrapped, I recognized how many genres I had no idea existed, which prompted my interest in understanding the potential of visualizing genre metadata to also have a clearer picture of possible movements in popular music.

Inspiration

My first step, before diving completely into the data, was to study existing investigations and visualizations. I was more intrigued by the evolution of the genres introduced by Spotify, which quickly led me to the incredible variety and complexity of research shown in Every Noise at Once. It uses an extremely complex algorithm to scan and categorize music at the molecular level, shown on a scatterplot of 5,782 genre-shaped correlations charted by Glenn McDonald, Spotify’s data alchemist. If you’re curious in what a genre like “Art Rock,” “Big Room” or “Permanent Wave” sounds like, you can listen to a sample by clicking on it or you can see this playlist made by Spotify.

Getting engrossed in Kwinten Crauwel’s interactive map of popular music genres and learning a lot more about it by watching his TEDx talk, A Visual Way to Explore the Musical Universe positively influenced my interest in creating a visualization to show genre popularity. One of the things that surprised me the most was how the map actually created after ten years of research and development and it not only ties subgenres but also their counterparts in other “super-genres.”

Materials

Kaggle: I found my raw data and downloaded it here.
Microsoft Excel – Spreadsheet software used to explore data and make a few basic edits.
Tableau – Free software to create interactive data visualizations to be shared in both public and private settings.

Process

1. Selecting and refining the data

With so much free data available online, I was having trouble deciding which dataset to utilize. The files were accessible in a variety of forms, but I decided to download a CSV file. Then I downloaded the Spotify – All-Time Top 2000s Mega Dataset spreadsheet from Kaggle, I proceeded to exclude some unnecessary information that seemed irrelevant to the project. For example, there was a lot more information from the years before the 2000s that I didn’t believe was needed in the dataset for my project.

2. Data Visualization Using Tableau Public

After the refining process, it was time to import the refined file into tableau public. It took me some time to get used to the software’s Functionalities because it was my first interaction with it. I spent a good amount of time trying to explore various features and possibilities within the system. That’s when I realized the challenge I actually experienced while categorizing the subgenres inside their parent genres. Without this step, it was getting increasingly difficult to visually digest the information. There were 116 distinct genres represented in the original dataset. Initially, I began to group them by region, as there are genres mentioned in the data such as Australian Americana, Australian Dance, Australian pop, Australian rock, Dutch hip-hop, Dutch Indie, Dutch Metal, and so on. That wasn’t working effectively as well as I had anticipated; it appeared dysfunctional and tricky to comprehend. Instead, I opted to organize the genres according to their parent genres. To do so, I had to first grasp the subgenre in order to appropriately organize them. Because Spotify has established so many genres, it is not always easy to recognize its parent category.

3. Color decision

I decided to designate a completely distinct color for each genre. Initially, I intended to create a custom color palette by inserting XML code into the preferences sheet. After spending a lot of time trying to change my Macbook settings and other ways that were not enabling me to edit, I reverted back to utilizing the default palette. I decided to use the same color for the same variables to create consistency while also allowing the audience to readily compare the data from each graph. Furthermore, I included an action to the map that allows the audience to click each genre to highlight throughout the three graphs. To represent the top artists and their genres, have used a light tint of color with a smaller area for lower values in the range and a dark tint with a larger area for higher values. This is because we tend to link lighter colors with lower density or lower numbers in a range, and darker colors with more density or higher numbers.

Results and Interpretation

Once everything was in place, I chose to show the top five genres in terms of their popularity and planned out the visualizations in the manner that I believed would best portray this data. Visualizations that displayed all of the genres individually were frequently too vast to read on their whole and needed scrolling. My technique involved investigating several relevant variables and then attempting to visualize them in order to uncover patterns (if any). Overall, 12 sheets of visualizations were created in this exercise, with one viz usually leading to the next in explanation in the subsections presented.  As a pattern appeared, so did some inspirational insights, I scaled back those 12 pages to four that I believed best reflected the results. These have been added to a Dashboard.

I chose to convey a narrative through the succession of information visualization graphs, with each graph delving a bit deeper to reveal the specifics of the genre’s popularity. The first chart highlights the evolutionary process of genre rise and fall in popularity throughout the years (the top 5 most popular parent genres). This chart shows that pop was the most popular genre, although other genres had their peaks as well, with Rock peaking higher than pop in 2016. We may also view the value by hovering the mouse over the line. To demonstrate how various popular genres have changed through the chosen timeline. I was particularly intrigued by the year-based bar graph results where  I wanted to highlight the specific subgenres that contributed to the general popularity of the genre for that year in the following image. The Gantt bar gave an intriguing insight as well, since it’s compact and easy to read and clearly depicts the sub-genres and compares them to demonstrate which sub-genre was the most popular. Furthermore, by dragging the year bar in the Top Artists and their Genre. I wanted to demonstrate the movement in popularity of genres and the musicians that represent them. We can notice the genre trend as well as the artist of that subgenre at the same time.

Reflection

Mapping out the top music genres of a decade provided an intriguing way of coming to terms with the complexity of how many genres actually exist. I learned so much about the subject that I picked, and it became a great experience for me to work on! This assignment taught me how much information is required to build even a basic line graph. I had a lot more fun using Tableau than I expected, and after a lot of practice and trying out a few different things, I started to notice a pattern and got more control in dealing with the data to create the representations I wanted. Learning how to use Tableau may be simple, but because I’m not used to dealing with data, it took me longer to grasp the technical parts of the application.

Future versions of this project might investigate other functions by integrating platforms other than Tableau Public. It would also be fascinating to try to correlate genre data published with other datasets of notable artists and groups responsible for the popularity of the top genres. Furthermore, diving into any accessible statistics on genre popularity over time, paired with country-specific data, would be interesting.