Charles Roberts
March 6, 2022
The Free Music Archive (FMA) is a website that offers free access to open licensed, original music which is curated by netlabels and thousands of independent musicians around the world. The FMA website claims that it receives tens of millions of visitors every month who download music for personal use. Users share and remix music from FMA in videos, podcasts, films, games, apps, and school projects.
The FMA was founded in 2009 by radio station WFMU-FM which is a listener-supported, non-commercial radio station broadcasting in Jersey City, NJ. WFMU is currently the longest running freeform radio station in the United States.
We live in an era of user-generated online content with the rise of YouTube, TikTok and Instagram. I thought it would be interesting to analyze an independent music platform, like FMA, because community-generated or user-generated content has the potential to impact a wider audience than branded content. A survey from 2019 states that “Consumers are 2.4x more likely to say user-generated content is authentic compared to brand-created content” Stackla, 2019. When I visit the FMA I get the feeling that the content is authentic and trustworthy as opposed to the Spotify or Apple Music brands where they are trying to market or sell something to you.
I used Tableau and Tableau Prep for data exploring, analyzing and creating visualizations. Using the FMA data, I want to answer the following questions:
Data Source: https://github.com/mdeff/fma
The FMA dataset provides data on 917 GiB of Creative Commons-licensed audio files from 106,574 tracks from 16,341 artists and 14,854 albums, arranged in a hierarchical taxonomy of 161 genres. The dataset was published by the International Society for Music Information Retrieval Conference (ISMIR) in 2017 and contains FMA data from 2009 to 2017.
There are 9 datasets included in the FMA GitHub repository but for my purposes I am focusing on select metadata from 2 data tables on the following fields:
tracks.csv
genre.csv
The data I am interested in analyzing was organized well so there was not much cleaning needed. I used Tableau Prep to clean up the column headers and remove unwanted columns. I create a calculated field in Tableau Prep to find how long each individual artist has been active on the website. I used the artist_date_created column which provides a date for when each artist became active on the website. Since the data table includes data up to and including 2017, I decided to create a DATEDIFF calculation between artist_date_created and 2018-01-01.
Artist Duration Calculated Field
DATEDIFF('day',[artist date_created], #2018-01-01#)
My initial question when looking at the data is who uses the FMA? Initially, I wanted to find the popular genres on the FMA but I found that many artists categorize themselves by multiple genres for their music. It was difficult to analyze the top genre based on artist “listens” or “interest” data because of the overlap of multiple genres for each artist. I felt it was best to look at the number of tracks associated with each genre.
I analyzed genre title and the number of tracks for each genre which shows that Experimental, Electronic and Rock genres have the highest number of tracks on the FMA compared to the other music genres. These results suggest that the Experimental music genre is predominant on the FMA.
I want to know who are the popular artists on FMA, but this can be a tricky question because there are a lot of ways to measure whether an artist is popular or successful. For my analysis I measured popularity on the FMA by listens and interest counts of tracks and albums.
The above bar chart shows the artists with the largest coiunt of track listens. The chart shows that Podington Bear has the largest count of track listens with a count 7,287,253.
The above bar chart shows the artists with the largest count of track interest. The chart shows that Podington Bear has the largest count of track interest with a count 9,922,230.
Counting the number of listens and interest is a good indicator of the popular artists on FMA but I also wanted to measure how long they have been active on the website. If an artist has been active on the FMA for a long period of time they would have more time to gain listens and interest.
The above chart shows the number of days the top 5 most popular artists, by listens and interest, have been active on the FMA. I used a calculated field to get the artists length of time, in days, on the FMA. I calculated the difference between the artist_date_created field and a chosen year of 2018 because the dataset only has data up to and including 2017.
Podington Bear has been on the website the longest, at 2814 days, which may be a good indicator of why they have the most listens and interest. Kevin Macleod, Chris Zabriskie and Jahzzar have been on the website for less time but they have similar listens and interest counts to Podington Bear. It should be noted that Kai Engel has only been on the site for 1600 days at this point but ranks quite high for listens and interest suggesting that they are a popular artist compared to the other top artists.
Podington Bear
Chris Zabriskie
Kevin Macleod
Jahzzar
Kai Engel
I thought it would be good to start off by looking at what types of albums are on the FMA. The FMA categorizes album type into five categories:
The above bar chart shows the types of albums ranked by the count of album listens. The chart shows that the album type Album is quite predominant with the most album listens over the other types.
I want to know what are the popular albums on the FMA. For my analysis, I compared the album listens count for all the albums on the FMA.
The above bar chart shows the albums with the largest count of album listens. The chart shows us that the album “Entries” has the largest count of album listens with a count 495,429,777. The “microSong Entries” album is the second most listened to album with 100,934,450 listens followed by “Bonus Beat Blast 2011” with 58,985,533 listens.
It should be noted that the “Entries” and “microSong Entries” are compiled contest albums where artists submit tracks that get judged for prizes. These contest albums are promoted on the FMA thus drawing a lot of attention and interest which may increase the album listens count. “Entries” features 108 artists with 139 tracks and “microSong Entries” features 115 artists with 310 tracks. The fact that these contest albums are promoted on the FMA and have many artists and tracks are good indicators as to why the album listen counts are large. The “Bonus Beat Blast 2011” album features 31 artists with 73 tracks and may possibly be the most popular album that does not have over 100 artists and is not part of any FMA promotions or contests.
Expanding on my analysis of the most popular album on the FMA, I looked at albums released by individual artists. For my analysis I measured each album’s popularity on the FMA by comparing the album listens count for each artist.
The above dot strip plot shows that Podington Bear is ranked the highest with 40 albums and 6,356,117 album track listens. Since we’re looking at individual albums, we can clearly see that the “Nameless: The Hacker RPG Soundtrack” by BoxCat Games is the most popular album with 1,533,769 listens. It should be noted that the album is a soundtrack by independent video game developer BoxCat Games featuring a compilation of mutiple artists. The data shows us that “Cylinders” by Chris Zabriskie is the most popular album by an individual artist with 1,363,291 listens.
Entries
microSong Entries
Bonus Beat Blast 2011
Nameless: The Hacker RPG Soundtrack by BoxCat Games
Cylinders by Chris Zabriskie
I want to know what are the popular songs on the FMA. For my analysis I measured a song’s popularity on the FMA by comparing track listens and interest counts for each song.
The above dot strip plot shows the songs with the largest count of track listens. The chart shows us that “Happy Birthday” is the most popular song but it should be noted that 23 different artists have this song title. A lot of “Happy Birthday” tracks are featured on the “Entries” contest album which would increase listens. The most popular song measured by track listens is “Night Owl” by Broke For Free with 543,242 track listens. The second most popular song “Starling” by Podington Bear with 491,234 track listens followed by “Springish” by Gillicuddy with 486,163 track listens.
The above dot strip plot shows the songs with the largest count of track interest. The chart shows us that the most popular song measured by interest is “Night Owl” by Broke For Free with 3,293,557 track interests. The second most popular song “Enthusiast” by Tours with 1,991,344 track listens followed by “Siesta” by Jahzzar with 1,563,555 track intersts.
Overall, “Happy Birthday” is the most popular song on the FMA but the most popular song by an individual artist is “Night Owl” by Broke For Free with 3,293,557 track interest counts.
Night Owl by Broke For Free
Starling by Podington Bear
Springish by Gillicuddy
Enthusiast by Tours
Siesta by Jahzzar
I want to know if there is a correlation between listens and interest for FMA artists. For my analysis I measured the correlation between FMA artists’ track listens and track interest counts. It should be noted that for this analysis normalizing the data did not alter the results.
The above scatter plot shows the correlation between track listens and track interest for each artist. The plot shows us that there is a correlation between listens and interest for artists that have under 1 million track listens and interest counts. As the counts gets larger most artists exist around the trend line. The popular artists sit above the trend line with larger track interest counts than track listens. Overall, there is a correlation between listens and interest for the majority of artists on the FMA because as track listen counts increase so does the track interest count.
I want to know if there is a correlation between listens and interest for each song by the popular artists on the FMA. For my analysis I measured the correlation between FMA artists’ track listens and track interest counts for each song.
The above scatter plots show the correlation between track listens and track interest for Podington Bear, Chris Zabriskie, Jahzzar and Kevin Macleod. It is interesting to note that there is a strong correlation between listens and interest for Chris Zabriskie’s songs which are distributed along the trend line.
I want to know where the popular artists on the FMA are geographically located. For my analysis I used latitude and longitude coordinates to position all the artists on a map. I used a size label to represent track listens in order to differentiate the popular artists.
The above symbol map shows that the majority of the FMA artists are located in the United States and Western Europe. It should be noted that I can not account for artists having a city, state/province or country location listed that is different than their listed longitude and latitude. Also, I cannot account for some artists who have longitude and latitude coordinates that position them in the middle of the ocean. I am not sure how this information is acquired and I decided to not alter this data in the cleaning process.
The above symbol map shows FMA artists located in the United States. The majority of artists I found in my analysis that I demonstrated to be popular are located on the west and east coasts of the United States.
The above symbol map shows FMA artists located in Western Europe. The map shows us that there is a large FMA artist base in Western Europe, but the majority FMA artists that I demonstrated to be popular are located in the United States.
The Experimental genre has the largest count of tracks on the FMA. Based on track listens and track interest Podington Bear is the most popular artist followed by Kevin Macleod, Chris Zabriskie and Jahzzar. The most popular contest album is “Entries” and the “Nameless: Hacker RPG Soundtrack” by BoxCat Games is the most popular soundtrack album. “Cylinders” by Chris Zabriskie is the most popular album by an individual artist. The most popular song on the FMA by an individual artist is “Night Owl” by Broke For Free with the largest count of listens and interest.
Track listens and track interest are good indicators of popularity on the FMA. I found that there is a correlation between track listens and track interest for artists that have under 1 million listens and interest counts. The popular artists have a larger number of track interest counts than track listens. The correlation between listens and interest is present for the majority of artists on the FMA because as track listen counts increase, track interest counts increase.
I found the majority of the popular FMA artists are located in the United States along the west and east coasts. Western Europe has a large FMA artist base but the popular FMA artists are located in the United States which is most likely due to the fact that the FMA was created by an American based radio station.
https://github.com/mdeff/fma
https://freemusicarchive.org/
https://stackla.com/resources/reports/bridging-the-gap-consumer-marketing-perspectives-on-content-in-the-digital-age/