I did two things related to music within the past few weeks: first, when Spotify launched in the United States, I immediately signed up for the Premium paid service.  I also read the book Electric Eden: Unearthing Britain’s Visionary Music, by Rob Young of The Wire magazine, a thorough and idiosyncratic look at British folk music.

Electric Eden covers a style of music I don’t know much about.  I know a handful of Fairport Convention albums, and I have a passing familiarity with the Strawbs, Pentangle, and Steeleye Span, but that’s about it.  The book made me want to explore this genre of, as Young calls it, “visionary” or “secret garden” music.

Not discover, but explore.

So I dove into Spotify and started exploring artists such as the Watersons and John Martyn, and the solo work of Sandy Denny and Maddy Prior (lead singers of Fairport and Steeleye respectively).  This process worked to an extent, but it was neither efficient nor particularly satisfying.  To start with the obvious: the year given with each album was often that of some CD reissue, not the original, leaving me with little idea of the artist’s chronology.  And the list of “related artists” that these services show you say nothing about how they’re related.  This is particularly tricky with genre-blurring artists like Martyn and Fairport.

One of the biggest buzzwords in today’s digital music scene is “music discovery.”  Music discovery simply means listening to something you haven’t heard before; or at least its meaning has been reduced to that lowest common denominator.  The principal tool for music discovery in most of today’s digital music apps is the playlist.  For our purposes, there are three kinds of playlists: fixed, variable, and shared.

Fixed playlists are things like “stations” on Live365.com or the “top tracks” lists on artist pages in Rhapsody and other apps.  They’re just artist or genre anthologies, or they’re tantamount to Billboard charts with click-to-play links.  You know what’s in them, or at least you do once you’ve listened to them.

With variable playlists, you don’t know what you’re going to hear next.  Broadcast and satellite radio are variable-playlist media.  Variable playlists are generated by humans (as on radio) or automatically (as on services like Rhapsody), and sometimes you can customize them to your tastes, as with Pandora and similar apps.

The latest trend is shared playlists, where users can create them and publish them to the rest of the service’s user community.

Service developers like these types of playlists in increasing order by how I’ve listed them here, because they have increasing amounts of that quality so dear to startups and their venture investors: scalability.  That is, you can create more of them without much (or any) additional cost.  To use another silicon valley buzzword that’s currently teetering on the brink of cliche-dom: shared playlists crowdsource music discovery.

But how much do playlists help with music exploration?  Not so much.

The ultimate shared-playlist capability nowadays is Spotify’s.  Spotify’s application interface not only lets people create and publish playlists within the app, it lets them publish Spotify playlists anywhere at all.  There are now several different websites, like sharemyplaylists.com, that aren’t affiliated with Spotify but provide Spotify playlists.  At this writing, sharemyplaylists.com boasts over 50,000 of them.  At least some of them appear to be supplied by record labels.

This may be “scalable,” but it scales into a mess.  Abandon hope all ye who enter sites like sharemyplaylists.com.

What I’d like is someone sitting over my shoulder and telling me “if you like this song, try X.”  Or some way of helping me navigate from one artist or song to another based on specific characteristics.  The Music Genome Project (MGP) has this capability but, I’d have to say, keeps it from reaching its full potential by locking it up in the enormously popular Pandora music service (yes, I’m a paying member).

MGP creates lots of information — or in techie parlance, metadata — about relationships among songs.  Tim Westergren, the founder of MGP, describes it briefly here; in addition to the items in his description, MGP stores metadata about band members, so that Band X is recommended if you like Band Y because they had the same drummer, and so on.  Trained musicologists create the metadata in MGP the hard (i.e., non-scalable) way: by hand.  That’s why there are “only” about 800,000 songs in Pandora’s library, compared with over ten times that many in iTunes, Rhapsody, or Spotify — or other “custom radio” services like Last.fm and Slacker.  The latter boast of larger collections than Pandora, but there’s a reason why Pandora’s subscribership beats the others: it’s the quality of the metadata.

If MGP were to “open” its metadata scheme and let users (and app developers) navigate it according to relationship types, it would become more of a music exploration tool.  I like that song; let me see what other bands the guitarist played in.  Or: I like that song, let me see other downtempo ambient electronic music in major keys released during the same time period.  Or: I like that piece of chamber music, let me see what other woodwind quintets were written during the Baroque period.

Another type of tool that would lend itself well to music exploration is the “family tree,” as epitomized in Pete Frame’s series of books such as Pete Frame’s Complete Rock Family Trees.  Imagine a tool that turned Pete Frame’s family trees into clickable graphics that helped you explore a band like the Velvet Underground or a genre like Grunge.

Or, how about a tool that takes the text of a book like  Electric Edenor the one I’m reading now, Paul Stump’s The Music’s All That Matters: A History of Progressive Rock (or, for that matter, blog posts and so on), analyzes the text, and creates not only a bunch of playlists for use in your favorite music app but also deduces relationships among the songs, albums, and artists.

There are lots of things that music services can do to make it easier to explore, not just discover, music.   It should be the obvious next thing to do in digital music services… if someone can find ways to make it scalable.

Advertisements