How People Discover Music

I missed it in July, but today The Listenerd drew my attention to a survey conducted by mp3 blog aggregator, the Hype Machine which asked people: How Do You Discover Music?

It doesn’t take a Ph.D. in social sciences (I hope!) to spot some immediate reasons to question the survey’s validity — it was conducted at an mp3 blog aggregator, which ought to skew it toward mp3 blog readers in a big way, and it was not open-ended (apparently they didn’t include “radio,” as was noted in the comments — 7.6% said they used the radio in the ‘other’ category, but would more have said so had it been asked?).

At any rate, though, here are the results of the 1,430 answers they got (people could check all that applied):

Total Responses: 1430


These findings have some interesting implications, though I think we have to discount the fact that “online editorials” (i.e. mostly blogs), came out as the first given the nature of the sample.

(1) Friends rule. Still. In their own interpretations, they go right to, but I am not convinced — especially after seeing the role of music in friending on in my survey data (still working on that stuff) — that it’s really online friend connections where most meaningful recommendation is happening. If it is online friends, I think most of those friends, or at least a very substantial proportion of them, are friends offline with whom one also connects online rather than friends formed online for purposes of sharing musical taste.

(2) Algorithms such as and Pandora’s (“Online Mechanical”) have a long way to go. This may be on account of people not using them, but I suspect it also has to do with the difficulty of capturing just what it is that makes Band A appeal to someone when Band B doesn’t. has spent the last 2 years recommending me people I know I don’t like, to the point where I quit checking my recommendations. Lately though it seems to have gotten somewhat better (except for that it now recommends bands I already listen to!).

This really suggests the tremendous potential of things like iLike and applications on Facebook for spreading the word. If you get the music recommendations where the offline friendship networks are already mirrored, word spreads. This has been happening on MySpace already for a long time, but people have acted as though it’s a straight band to fan phenomenon when it’s always been friend to friend, with the band as a social commodity exchanged between them.

It also means that the more ways that labels and bands can make their music portable so that friends can ship it around easily to one another, the more they’re going to thrive.

It was funny to run across this because just yesterday I was thinking that someone really ought to do a study where they ask a large random sample of people how they heard about the last artist whose music they got into and whether or not they had purchased the music. That’s what you’d really need to do if you wanted to know how people are getting turned on to new music. Unless someone wants to hire me to do that study on their behalf, I doubt I’ll ever do it myself, so please steal this idea (or hire me!) because I want to know the results.

Comments (2) to “How People Discover Music”

  1. I’m lucky enough to live in a city with a great community radio station, and long before the Internet I lived within a few miles of a great college radio station.

    A lot of times my Last FM recommendations are bands I already know and listen to, also. Some of them are incredibly obvious, even if you slide the bar to “more obscure.” (Neko Case and Her Boyfriends if you like Neko Case. Er, that’s the same artist with a backing band.)

    These days I get a lot of good recommendations from other people on the fan boards I frequent.

  2. I think that you are looking at only from the algorithmic perspective. That is to say, only from the view of what the “machine” tells us we may like. In its infancy,’s recommendation feature was far more prominent than it is today. I think that they have relegated it back to another page is exactly what you have stated.

    However, the importance of has far more to do with the human interaction than the one based on a computer model. For example, when I first visited your page your plays and mine were not even close, and thus we would probably not refer bands to each other in the algorithmic sense. However, like in the real world, I was able to read your journal entries and found that you share both a love of some of my favorite bands of the 80’s, but also a passion for music that is bordering on obsessive (which in my world is quite a compliment).

    For me, I scan people that sound interesting and pluck several of their artists to determine if I’ll like them. I have never met any of my friends (save my sis-in-law of course) but I would think that many of them could be friends offline based on music alone. This is truly fascinating and the metrics are certainly daunting to control for outside variables. I am expecting a whopping new book on the subject from you! Haha. If you ever need help with it, count me in!