Pandora: The More You Listen, The Better It Gets
Whenever a user likes a new song, Pandora’s playlist choices improve for ALL uses
Pandora is a platform where users can pick a station (linked to genre or artist) and immediately start listening to music. Users are encouraged to let Pandora learn their taste by “liking” and “disliking” the songs it picks for them.
The direct network effect is a strong one. Practically all users give Pandora feedback about the songs it picks – likes, dislikes, skips etc. – and this does two things: (1) it allows Pandora to understand each user’s taste better and (2) it improves its understanding of its songs database, thus helping all its users. The first allows it to know what song it should play next; the second helps it to decide what songs it should play to all its other users. Let’s suppose, for example, Alice just liked a Taylor Swift song and a Katy Perry song. If now Bob likes a Taylor Swift song, Pandora is more likely to pick Katy Perry for the next song for Bob. This author was definitely impressed by how quickly Pandora “figured him out” and consistently started introducing him to new songs he likes. If you’re looking to discover new music with a low miss rate, then in this author’s opinion, Pandora is your best chance.
This spells bad news for anyone trying to challenge Pandora. With so much data collected, and experience in making use of it, a challenger would have a hard time replicating that part of the user experience.
At the moment, Pandora captures value, unsurprisingly, via (audio) ads. With over $900M in revenues in 2014, Pandora is expected to turn profitable soon (last year it lost $30M, or 3% of its revenues) – as reflected by a market cap of over $4B. This author believes that value capture could improve going forward. First, as Pandora grows, its negotiating power improves when striking content deals – with more users, content owners will be forgoing a major revenue stream if they refuse to partner (and interestingly, unlike Spotify, Pandora can afford not to include many artists in its offering, as it’s not an on-demand model, so users won’t be let down if an artist isn’t featured on Pandora). Another way for value capture to improve is if Pandora partners with Facebook or Google, and gets them to add audio ads to their offering. Google and Facebook would know what user should listen to which ad (raising effectiveness, and thus advertisers willingness to pay), and Pandora will execute. Both Google and Facebook are already offering third-party apps and sites such a partnership, so this seems like a natural continuation.
Another evolution could be towards making Pandora a platform for the music industry to test market demand for a new artist or song. With so many users lending Pandora their ears, and providing feedback, Pandora is well positioned to do the job traditionally done by record labels. Today, the labels scout talent, and try to make an educated projection of the likelihood of an artist making it big. With Pandora, one could test demand by inserting a song to listeners’ playlists, and measuring their reaction. One could even easily A/B test. This would add two parties to the Pandora platform – providers of content to be tested (upcoming artists, or established artists looking to test a specific song), and providers of capital willing to fund the testing. This would systemize music creation to best target actual market demand.
There are many ways Pandora could grow its business, but what’s almost certain is that we’ll get to enjoy better content and more personalized picks going forward.
Great post, Asaf. I think your point about Pandora serving as a platform for new music and A/B testing is an interesting one. Do you think Pandora is better positioned than applications like SoundCloud to pull this off, speaking more specifically about SoundCloud’s social footprint and large user base.
SoundCloud is an app where new music is listened to, liked, commented on, and shared across millions of users. They show the exact second or place in a song where a comment is posted. Talk about relevant, real time feedback!
Thanks for your comment, NoSocks!
Hmm… embarrassingly, I was not familiar with SoundCloud to be honest. This does look like a very good platform to get an artist some real time feedback I must say. One thing I would be concerned about, if I had to think of one, is that maybe the SoundCloud audience isn’t representative enough? I’m guessing it’s the people looking to find new and obscure music have a different taste than 99% of music listeners out there. Again, not a real concern, but something to keep in mind.
I think the comparison with SoundCloud is interesting, though to me the two models are slightly different and the value of network effects is different as well. Pandora in a sense curates playlists for its users by trying to best replicate their tastes and preferences. As you describe, network effects help the platform to do this better by collecting data on users’ reactions to its choices of songs. SoundCloud, on the other hand, is more hands-off as a platform. Artists themselves choose to upload music, and listeners find, make and share playlists, tags and comments. The value of network effects here seems to me to be more around building up a community whose scale makes it attractive – the more users there are, the more artists will want to participate, and vice-versa. So while Pandora uses network effects to improve the quality of its service, SoundCloud seems to use them more to improve the quantity/scale of its content. Of course SoundCloud’s tagging system allows users to explore new music based on what they’re listening to, and to see what’s trending, but I still think that unlike Pandora, whose main value proposition is to actively suggest new music to listeners, SoundCloud’s primary value is in having an active community of both artists and listeners. Both seem to be working well – it’ll be interesting to see how both evolve!