Discovery Without Algorithms | July 2013

I have a confession: I hate Pandora Internet Radio. I feel like that is confessionworthy since it is the most popular service of its kind and people clearly love it. It’s not that I dislike the service, or other similar services like or Spotify, because I have anything against streaming music. But I don’t trust Pandora’s music discovery capability, which is what drew many people to the service in the first place.

When I first used Pandora’s music discovery feature, referred to as the Music Genome Project, I was genuinely excited. I thought that the creators had come up with something amazing: a math-based system for figuring out what music I would like. The system analyzes 400 musical attributes, or “genes,” such as gender of lead vocalist, time signature, and level of distortion on the electric guitar, in order to find recommendations. Each song is represented by a vector, or a list of attributes. Rock and pop songs have 150 genes, rap songs have 350, and jazz songs have approximately 400. Other genres, such as world and classical music, have between 300 and 500 genes. The system uses algorithms to generate lists of songs with similar vectors.

Starting my search

I typed artists’ names into the search box and listened to song after song, my cursor hovering over the “thumbs up” and “thumbs down” buttons, which is how the user helps Pandora refine the selections. But I found myself using the thumbs down button much more often.

If I happen to like a female singer who played guitar on a song in a minor key, it does not necessarily mean I’ll like anyone who simply matches that description. What’s missing from the vectors are the indefinable qualities of a great song, such as the unexpected brilliance of a key change or the cadence of the singer’s voice. After trying the service a few more times, I gave up, dejected that math had failed to tell me what I should like. I still wanted to find new music to love, though. I longed for the excitement of hearing something for the first time and getting goose bumps (something that has been scientifically proven to happen).

Days of new music

I decided to listen to a new band every day during 2013. I thought about my favorite musicians and how I came to discover them for myself, realizing that I often found out about them because of their work with musicians I already knew. So I turned to the events calendars at music clubs around the country and even the world. For example, I’m a fan of  The Frames, an Irish rock band. Glen Hansard, The Frames’ lead singer, collaborated with Czech musician Markéta Irglová, with the duo calling themselves the Swell Season (you may know them and their music from the movie Once). I saw them perform at Radio City Music Hall with Josh Ritter and the Royal City Band as the opener.

Through them, I’ve found the David Wax Museum, who led me to Pearl and the Beard, which led to both Lucius and Lady Lamb the Beekeeper. As this line continued, the venues got smaller, but the joy this music brought me did not.

The headlining bands who invite openers to go on tour with them are providing a similar service to Pandora but on a more personal level, in effect saying, I like this band and trust them enough to set the tone for the night, so maybe you’ll like them, too. You don’t even have to go to shows to do this. Start with a band you like and check its website to see who is playing with them, then check out where else those opening bands are playing and who is on the bill, and you’ll quickly find new music to love.

For my project, I tried focusing on musicians whom Pandora would invariably miss, many offering free downloads on websites like I shied from bands that already had a strong social media presence, looking for those that did not have tens of thousands of likes on Facebook or a legion of Twitter followers. As I began to compile my list, I decided to share what I listened to on a daily basis, using the hashtag #365daysofnewmusic on Twitter.

Looking ahead

After barely half the year, this project has already been a success. I’ve found new bands that I love and found other bands by going to shows. Word of mouth is essential to finding new music, and it’s something I do as a reference librarian, offering recommendations to patrons based on the CDs they are borrowing. Music discovery websites have taken note, offering a more personal approach. Some of these services use playlists and a voting system to provide recommendations, while others use algorithms based on user choices, similar to Pandora but without the more scientific breakdown. They are also heavily pushing the social side, making the apps more like Twitter or Facebook, where you can follow other people and interact with them through the app. But, really, nothing beats standing in front of a stage, not knowing what to expect, then feeling goose bumps rise on your skin. An algorithm can’t do that for me yet.


Brian Morell is a Reference Librarian and Assistant Manager, Queens Library, Jamaica, NY. Follow him on twitter @goodinthestacks to keep track of all the bands he’s listening to for #365daysofnewmusic



  1. Jen Stayrook says:

    I had to give up Pandora for this reason as well. That and the radio had ten or so songs it REALLY enjoyed playing. Over and over and over.

    I much prefer Spotify for finding new music. The related artists tab is great and once I go to an artist’s page, I can see their top tracks; usually the top songs are an indicator if I’d like the music.

  2. Kasemopolitan says:

    I love that someone finally wrote about this !