I consulted on such a product a while back and unfortunately the task is much harder than you might think. The project I worked on was actually easier than this because what we were doing is taking a bunch of the local unsigned bands in Los Angeles (where I live) and trying to use listener preferences to introduce users to those bands. But there were a few significant hurdles...
1. Identification of music is quite expensive. It would be great if a company like Pandora would license their database but until that happens there's a major financial burden on whoever has to pay people to identify existing music in order to make an engine like this work.
2. It's hard to determine user's likes. One of the biggest and most annoying flaws in iTunes is that its "times listened to" count doesn't consider a song listened to until you get to the very end. So if you (like most people) hit "next" when you get to that point at the end of a song where they repeat the lyrics over and over again than iTunes doesn't count that. Which makes it hard for another program to determine your preferences.
3. People have a low tolerance for new music from an artist they don't know. Ideally you want to slip a new song in amongst the listeners favorite songs. But that doesn't work when you have to write your own program (unless you have the money the pay royalties for mainstream songs). Which means you're stuck with trying to make the user listen to 20 new songs in a row which most won't do.
Again it isn't a bad idea if someone wanted to try to run with it but I'd consider the above before putting a lot of time into it.
Are you sure that "Most People" hover over the fast forward button like that? I've spent a lot of time around a lot of people with a lot of music playing, and have never once observed anybody actually doing that.
I actually did have a friend back in the day of cassette tapes who used to drive around his neighborhood until the music played off the end of one side before parking, so that his music would be correctly rewound next time he got in the car. I suspect he thought that everybody else did that too.
Spotify comes pretty damn close, if you are in one of the countries they support (UK, Norway, Sweden, Netherlands.)
For $10/month, you can listen to unlimited music while online, and offline on your iPhone or PC. For $1/per track, you can download songs to load to other devices. And, using the Facebook integration features, you can subscribe to Playlists (by your friends, or others) that will recommend music to you.
I wrote that post when I looked at three things:
1) how much money I spend on music each month (currently more than $20),
2) how much time I spend listening to music each month, and
3) how much time I spend looking for new music each month.
Then I realized I spend a fair bit of time listening to music away from my computer (commuting, at the gym, etc). I figured it would be nice if I could use this time to find new music. You see, when I buy music, it usually isn't because it's what I set out to do. It happens because I hear a song that is good enough for me to stop what I'm doing and investigate.
If my phone had the battery life that allowed my to use the last.fm or pandora apps more or less continuously, that might get me pretty close. (I should probably give this another try.)
But here is what I find myself doing... There are a number of sites that scan the music blogs for mp3 posts and condense them down to a single source. I see those (passively) in my twitter stream. This makes it easy to sample songs quickly in my browser. Ones that interest me get downloaded and end up in a special playlist in iTunes. This playlist syncs to my iPod. These songs get cleaned out and moved into another playlist once a week. About once a month or so, I go through this playlist and look for songs that I've given high star ratings too or have played >N times. I then go checkout those artists to see if I should be buying their music.
I've found a lot of good artists this way. If this seems like a lot of work, it probably is. I spend 10 hrs a day coding at work and I have more fun doing that when I'm listening to music I like, so I'm motivated to spend time and money on this problem.
Doesn't Zune Pass do that? If you live in the US it should work. For $14.99 you get music streamed, but also downloaded to your Zune, and you can keep 10 songs per month in non-DRM'd format. And SmartDJ should be able to do the automatic downloading (although I don't know how good it is).
(From Zune.net: "The songs are yours to play on up to three computers, and three Zune players and phones, for as long as your Zune Pass is valid." and "Additionally, a Zune Pass gives you 10 song credits to redeem each month. You can keep these songs forever - they become part of your permanent collection.")
This is the only device and marketplace that meets -EVERY- -SINGLE- request the author made.
It gives him his unlimited music. It gives him his DRM-free tracks to keep each month. It gives him his recommendations based on his tastes. It gives him his automatic download whenever the Zune connects to the internet.
Dear Music Labels and the RIAA,
This is the kind of cool stuff we could build if you guys weren't so draconian about licensing. This is a great idea that would lead to people listening to, and buying, more music every single month. But you would have us load it up with DRM, or make it so you can't skip songs, or some other bogus restrictions in your fear of anything new.
Conceptually simple, but the licensing agreements with the labels would kill you, not to mention the problems with divvying the money amongst the artists that you decide to "keep."
That brings me to another problem: you'd have to pay $25/month for this service. $5 for unlimited streaming (the Rhapsody way), $20 for 20 tracks at $1 each. Or you could do 5/15, or something like that, I guess.
It's a lot of risk for not-so-much reward, and as such is a bad investment proposition from where I'm sitting. If I were a VC, I probably wouldn't invest in such a company. It'd be much easier for an established player to make this happen. They have the infrastructure already built, the legal expertise to make it happen, and the connections in the industry already.
I've been working on something pretty close to this for some time now. The hurdles mentioned by TomOfTTB are valid, but like another commented, there are services in place. It's really about stitching these systems together, but also identifying who the real customer is.
The author is right:
we're going to need to find another model
Spoiler: the customer is not the listener. Think Topspin, Bandcamp. Combine that high level model to some of the features mentioned in the article, and you're getting close. I hope :)
I don't see why anybody would need a device like that. There's an offline option in Spotify for example, and there's hardly any way you could make the experience better than that, and it's only 4.99 a month. I don't know if there's currently similar services for US markets though. Probably Apple's going to launch something similar tomorrow.
1. Identification of music is quite expensive. It would be great if a company like Pandora would license their database but until that happens there's a major financial burden on whoever has to pay people to identify existing music in order to make an engine like this work.
2. It's hard to determine user's likes. One of the biggest and most annoying flaws in iTunes is that its "times listened to" count doesn't consider a song listened to until you get to the very end. So if you (like most people) hit "next" when you get to that point at the end of a song where they repeat the lyrics over and over again than iTunes doesn't count that. Which makes it hard for another program to determine your preferences.
3. People have a low tolerance for new music from an artist they don't know. Ideally you want to slip a new song in amongst the listeners favorite songs. But that doesn't work when you have to write your own program (unless you have the money the pay royalties for mainstream songs). Which means you're stuck with trying to make the user listen to 20 new songs in a row which most won't do.
Again it isn't a bad idea if someone wanted to try to run with it but I'd consider the above before putting a lot of time into it.