The Curator vs. The Algorithm

3 07 2015

A new battle for how music is distributed is now underway. While there are plenty of changes in the music business, one question is how to get a lot of music listeners without requiring they purchase their music. Generally lumped under the terms “streaming music” or “internet radio”, the idea is to get “music to the masses”. How this ultimately appears has yet to be determined.

The Ups And Down Of Distribution

Music is generally distributed through either physical media (currently CD and vinyl records), through digital download, through traditional radio (both broadcast and satellite) and through streaming services. The trends are intriguing:

The sale of vinyl is a bit of a curiosity. It would be easy to say it is confined to a niche of audiophiles and those interested in retro trends. But the volumes (now 6% of all music sales) are such that it appears vinyl may going beyond those small markets. Whether this is a short-term thing, or a real longer term trend isn’t obvious.

But the real battleground is streaming, which appears to be taking over the segment formerly dominated by radio, while reducing the demand for purchased music.

Historically We Don’t Buy Music

While the music industry today focuses on sales, the reality is that more music has been consumed through broadcast means than via purchases. Consider that the median CD collection size from the 1990’s was 30-50 CD’s. Most people simply replaced their vinyl with CD’s, so you can extrapolate that the media collection of records was similar in scope. Guys like me, who had a CD collection of around 200-300 discs, were the exception (and having an iTunes library of over 12,000 songs and growing makes me another outlier).

But on radio, we listened to hundreds of songs each week. Getting a song into rotation with serious airplay was a critical part of marketing to get people to add to their meagre collections. But it also got people to listen to the radio, which meant that the top stations could make a lot of money selling commercials. Listeners got to hear new music. Labels made money by taking a share of the ad revenue.

One of the common venues for listening to the radio was in our cars. But when cassette decks became more-or-less standard equipment, the decline of radio began. It continued when the cassette was replaced by the CD. Then MP3 players hit the scene, and now we could bring a huge library of songs into the car (and onto the bus, the train and into the office). Both the radio and CD suffered when services like iTunes came out, removing the need to buy the media and allowing us to make our own “radio stations” with playlists of our favourites without DJs, news breaks and commercials. But it presented a new challenge: finding new music.

Streaming Is The New Radio

Streaming services like Pandora, Spotify and Apple Music appear to be occupying the place that radio once held. It is the way we listen to old favourites plus find new music. Some, like Apple Music, feature “radio stations” with DJs. Others are basically just big playlists. Some offer free access in exchange for listening to commercials. Pay a monthly fee, and you can skip the commercials.

It means that listeners can have access to hundreds of thousands or even millions of songs for a modest monthly fee. It would cost millions of dollars to own a collection that size. Storing that much music would take terabytes of space, something that isn’t available on mobile devices.

Having a catalog of music and a service to stream it is one thing. The next is to get people to listen. The new battle is between two approaches to building those playlists: an algorithm vs. a curator.

The Algorithm

For most streaming services, a “channel” is basically a playlist that uses metadata on the song tracks to figure out if they belong in rock, pop, jazz, techno or what-not. The songs are played in more-or-less random order, with no thought as to what song “makes sense” after the current one being played. They may not even bother crossfading between tracks.

This is the algorithm. Songs are played based on data that describes the song. Some services will use sales data in other venues to figure out what should and shouldn’t be in the list, others likely take feedback from the labels as to what they want promoted. Others take listener feedback and preferences (for those that allow custom playlists) to put popular songs in the default channels.

The algorithm can make it hard for new artists to break through. It is, in some ways, like the current label/radio system, where labels work with the broadcaster to promote songs the labels want promoted. If CD/download sales data is included to rank what makes the list, then again, the label has a fair bit of control as to what makes it into the sales catalog and what doesn’t.

The Curator

The curation approach uses real people to select songs for these playlists. They may even influence what order the songs appear in. What goes into a playlist isn’t just whether it is tagged with the appropriate genre. It is about whether a song is interesting enough that the list curator feels people might want to hear it. Listeners provide feedback by making the song a “favourite” of some kind, and the curators use that to guide their search for other music that is similar.

The curation method has the potential to allow smaller artists to succeed without a label. Apple in particular appears to be setting up a system that, like the iTunes App Store, allows independent artists to get a distribution and promotion platform for their work. This, combined with social media to help with promotion, could allow small artists to sell their music without having to make something that appeals to a label.

The potential problem here is the same one that app developers have: getting people to find out about their app in a ecosystem of millions of competitors, and that comes down to marketing and promotion. In the music business, this promotion is the job of the label. They find music they think they can sell or promote, and their use their network of connections to build awareness.

The curator can help with that, by putting songs or albums they find interesting into their collections for their listeners. But they can also act as gatekeepers, and trying to impress a curator could end up being as hard as getting interest from a label. If a curator is having to review thousands of new tracks each week, that makes the job even harder. It may not be as big a boon for smaller and new artists as some think.

Which Will Win?

Which will prevail in the end isn’t clear. Both will likely coexist for a long time. But one may come to dominate. Right now, Apple is the only real player in the curation approach (although they have algo-based playlists driven purely by sales or popularity). If they do a good job, and the approach scales, then it may take the crown. DJs had some say in what go on the air, and Apple’s approach isn’t conceptually different from that.

But the algorithm approach scales better. It’s about the metadata and statistics. For better or worse, it does leave the labels in the driver’s seat. It also opens up the system to gaming, where artists pull tricks to fool the algorithms (similar things are tried periodically in the app stores to get an app onto a top-10 list).

The next few years should be interesting. The competition will lead to innovation and improvements, and we may end up with some kind of hybrid approach. How this works out will likely set the pattern for music sales and consumption for another generation.



%d bloggers like this: