When I was in middle school, I used to scour record stores and listen to radio stations and make lists of the songs I liked. Then I would give these lists to my friend, who would download them and burn me CDs. Our parents warned us that we would be in jail by the end of the week.
Once I got tired of a CD, I would rinse, lather, repeat. This was what “music discovery” meant to me.
Thankfully, things have changed. Predictive music services are one outlet for finding new music, and one that most of us enjoy using often. Much ado and investigation has been done into which really find you the “right” music, which stray from genre too swiftly, and which are perhaps too predictive, cycling you through all-too familar songs you’ve out-listened.
These platforms can talk about their engineering and algorithms and research and library catalogs all they want, and I could re-hash for you in laymen’s terms how they will generate a perfect playlist, and then we could all feel sort of placated and bored.
Or we could put them to the test: four streaming services. Four predictive players. One band to start with. 20 songs. Where will we end up? What will actually happen when you test the “same” station with different services that use different algorithms?
Let’s find out.
The idea here is to start with one band and put it through the algorithms of three different predictive algorithms and to see the twists and turns we’re taken on – as well as how much we end up liking the results. Let’s go over the experiment controls. And let’s get this out of the way: We’re talking about music here. There’s going to be a heavy dose of opinion; if you want to yell at me for not liking Broken Bells… well, I guess that’s fine, but don’t interpret this as a music review piece – it’s an algorithm review piece.
The band: Arcade Fire. The reason? I like listening to them and I’m going to be listening to a lot of this over the next few days – and also, they’re eclectic enough that we could go down some interesting musical paths.
The players: Spotify Radio, Pandora, Google All Access Radio, iTunes Radio
The songs: 20 songs, with all the natural thumbing up and down as needed. I’ll be tracking how many songs I like and dislike, as well as what new music I find, and how often the players throw in the control band’s music.
“Thanks to our all-new intelligent recommendation engine and multi-million track library, Spotify Radio is a music discovery experience without equal… It’s a combination of our own data based on billions of hours of on-demand listening from our users, and social data and thumbs data (liking it/saving track to playlists). I think what is unique to Spotify is that we are able to utilize internal listening data to inform our radio recommendations. ”
|We Used to Wait – Arcade Fire||Suck it and See – The Arctic Monkeys|
|Rest My Chemistry – Interpol||Maps – Yeah Yeah Yeahs|
|Une Anne Sans Lumiere – Arcade Fire||Lemonworld – The National|
|The Ghost Inside – Broken Bells||Down by the Water – The Decembrists|
|Under Your Spell – Desire||Lasso – Phoenix|
|Abraham’s Daughter – Arcade Fire||City with No Children – Arcade Fire|
|Codes and Keys – Death Cab for Cutie||Wasted Hours – Arcade Fire|
|The Day is Coming – My Morning Jacket||Sea Legs – The Shins|
|Rebellion (Lies) – Arcade Fire||Mongrel Heart – Broken Bells|
|Down River – The Temper Trap||Underneath the Sycamore – Death Cab for Cutie|
Honestly, I’ve had my doubts about Spotify Radio. I always felt like I was being pushed in the wrong direction. However I haven’t used it much since launch and I’m happy to say my original issues with it have been addressed. But that doesn’t mean Spotify Radio was able to deliver the perfect predictive playlist. If anything, it went too far the other way: Instead of playing music I felt was… well, wrong and outside the genre I was looking for, Spotify played it very safe.
Spotify Radio had more Arcade Fire songs in this playlist than the other two services, clearly trying to keep with the original station I picked. The National, Interpol, and Death Cab for Cutie, while bands I absolutely enjoy, are also the ones I always lump into my Arcade Fire playlists. These are obvious, safe picks – and it made for a fairly pleasant if boring listening experience. Also, it has to be noted that every time I hit thumbs down, Spotify just took me back to Arcade Fire. Overcompensate much?
The small ventures outside the “normal” were hit or miss: Things got a little too depressing emo-indie with Broken Bells, and the up-tempo part of Arcade Fire I like was definitely being lost with songs like Suck it and See and The Ghost Inside. Overall, Spotify Radio kept the peace and kept me placated but didn’t little in the way of discovery.
Thumbs up: 9
Thumbs down: 5
Songs I’ve never heard: 3
Arcade Fire tracks: 6