Disclaimer: I am the founder of Tweekly.fm, a service that automatically sends an update of your top artists for the week to Twitter and Facebook.
Recently, Spotify (with Sean Parker’s help) got engaged to Facebook. In short, if your Facebook and Spotify accounts are connected, you will see your friends listening to music in the new ‘ticker’ as they are listening to it. A lot of people have wondered, ‘Is this useful?’, ‘Why would I want to see John listening to Backstreet Boys?’.
Before I answer that. A little background. I started Tweekly.fm in Januray of 2009. The goal was simple. I wanted to make a twitter app. I was (and still am) an avid last.fm user, so I thought it would be great to automatically share your music tastes from last.fm to twitter. At that time #musicmonday was still big. People were sharing their music tastes on Twitter every monday. It was great! Where is it now? And what happened?
The biggest culprit is Twitter’s trending algorithm. They changed it to display only novel topics. In other words, because #mm was trending every monday, it wasn’t exactly novel each time. But why didn’t people continue sharing their music tastes despite this? There was no real return. During 2009 there was also quite a rise in websites that offered the ability to tweet your songs to Twitter. I wrote a blog post on this quite a while ago in April of 2010. It was an exciting arena, one in which Tweekly.fm was competing in as well. Of those sites in that blogpost (besides Tweekly.fm), only tweetmysong are above 450 000 in alexa rankings, and blip.fm remaining at the top (because of its built-in network effects). In short, the small ‘sector’ kinda died. Nobody took the effort to tweet a song they are listening to, because not a lot of people took the effort to listen to it. In other words, little return for both people. If someone shares a song with me, it works better if there is context. For me to like the song, there are two big prerequisites: If it is a good friend, who knows my music tastes, I will absorb the effort to listen to the song. However if it is an artist (and genre) I’ve never heard of, I still have to make up my mind about, because my friend shared it with me, and expects some return. In other words, I have to make an effort to form an opinion on the song. “Hey Simon! What did you think of Portugal. The Man?”, “Uh. It is great. I kinda liked the jazzy sections in the song New Orleans”.
If it is from a ‘musical’ stranger, the only context I have is if the person elaborates on the song. “Listen to Nero – Innocence. Epic dubstep in every way”. Now I know it is dubstep and if I am a fan, I would be more willing to accept the opportunity cost of taking the time to listen to it. However, for the person who shared the song, they still need a return. If I liked the song, I must still do more effort to tell the person that I liked it, and once again the interaction rate drops off heavily.
Why is Tweekly.fm still growing? It is automatic and it has context. There is no effort on part of the listener. They just have to consume their music and it will be shared each week to Twitter. The second factor that Tweekly.fm does to a certain degree is context. 3 artists are shared in the update. This means that if people see one artist they like in the tweet, they will be more inclined to click on it. If there are 2 known artists and one unknown artist, they will be even more likely to click on it.
So why is Spotify and Facebook on the right track? Music sharing works best when it is automatic, because it takes no effort on behalf of person sharing the music. They thus expect little, if nothing in return. Any comments on the artists you listened to is as they would say in marketing terms: a satisfying experience. Same goes for the consumer. They have no expectation to comment on the artists you share, but will be delighted if they find they share music tastes in common.
However, where their system fails, is music discovery. The only context being employed is the user listening to the music. If you know him to listen to cool electronic music tracks, you will be inclined to find out more AS they are listening to it. If you follow what they are listening to, you might pick up a pattern and then be inclined to look up the tracks yourself (“ooh, I know that song! oooh, I know that one as well! Oooh, I better check this one out, I don’t know it”)… But this I feel, is perhaps way too much effort. It beats the purpose of automatic sharing.
Automatic sharing allows serendipitous behaviour to arise, because of the non-effort to share it. In Facebook’s case, it doesn’t clog up the stream, because it occurs in the ticker. That is great. However, if they really want to ramp up music discovery, they need to use music recommendations to explain the context of songs that people are sharing. If Robert Scoble is listening to bluegrass band that I’ve haven’t heard, the system should preferably show context while he is listening to it. Like when Last.fm recommends new artists, they recommend it based on your current library of artists. In other words, it should preferably match up the closest artist I have listened, while also providing other information (such as genre and current position in world charts for example). This way, I can immediately discern context when music is automatically shared.
Who should be doing this? Last.fm. They have the resources and data available to do this. I can already see what my ‘friends’ on there are listening to, but there is no context. They know what I’ve listened to, they should just bring it together. I wrote a script the other day to test this. Of my 87 friends on last.fm, it returned to me the following dataset:
Of my friends who listened to music the current week and based on last.fm’s current music recommendations to me, I should listen to:
[Beirut] => 4 [Skrillex] => 2 [Björk] => 2 [Gold Panda] => 1 [The Wombats] => 1 [Band of Skulls] => 1 [Cut Copy] => 1 [Röyksopp] => 1 [St. Vincent] => 1 [Cults] => 1
It is very slow, because I have to make several API calls, so it is not available for testing (I might upload it github later). So in other words, what the above code says: “Of the artists we recommend you listen to, your friends listened to Beirut, Skrillex, Bjork, Gold Panda, Wombats, Band of Skulls, Cut Copy, Royksopp, St Vincent and the Cults this week”.
So: What it boils down to. Facebook and Spotify are on the right track. Music sharing works best when it is done ‘frictionlessly’, but now it just needs more context. I hope Last.fm gets there before them, but maybe it is just because I am a bit biased.