Monday’s NPR Marketplace featured a story on album song order – how it still matters even in the age of digital music. According to Billboard Magazine's Gary Trust, the evidence suggests that the earlier a song appears on an album, the more likely it is to be listened to. People listen to an album’s tracks in order, even digital music that can be purchased as individual songs. Listening to the story, I wondered whether in the age Big Data and digital music, analytics was playing a role in determining what we hear, when we hear it, and its order on an album.
Considering that analytics is being applied to everything from sports schedules to retail prices and flight planning, why not music? So I asked our frequent blogger Mike Farrar what his thoughts are on the topic…
1. Mike, does analytics have a role to play in determining song order?
Analytics probably could with enough data, but my guess is that people often approach an album like they approach a book – they start from the opening chapter.
2. Could this be generational … starting at the beginning?
That is definitely a possibility. Song order may come to matter less and less as consumers get used to purchasing individual songs online. We probably will tend to get further away from the metaphor of an album as a bundle of tracks to be played one after the other.
For now, though, even if the physical album is no longer the medium for distribution, it seems to me that the concept of an “album” remains relevant in this digital age. An “album’s” songs do tend to be similar to one another in quality and aesthetics, which means it’s worth a little investigation before committing to pay for that entire bundle of music.
You can probably get a pretty good picture of an entire album by listening to just a couple of its songs. You could build a fancy randomization algorithm to choose which tracks to listen to, but who has time for that? Just grabbing the first two songs is a simple, quick rule of thumb.
Here’s where it gets fun: if consumers are biased to judging albums on the first one or two songs, musicians have an incentive to game the system by putting their best two songs at the start of the album. A friend of mine assesses whether she’ll like an album by listening to its third track. She figures that any album can have a couple of good songs, but if the third song is pretty good, the album as a whole has a fair chance of being pretty good. It sounds kind of wacky, but she’s actually counter-gaming what the artists were doing in the first place.
The more interesting question is which specific songs an artist should put first, and which ones last. Now, a label could probably figure this out through focus groups and extensive research … but gut feel may actually be equally as effective.
Remember that just because you can get a mathematical answer doesn’t mean that it’s the best way of going about it. If artistic judgment is basically as good as rigorous analysis – and a heck of a lot easier and less expensive to implement – it’s probably a better way of going about things.
3. What role can analytics play in the music discovery process?
This is where analytics can really shine. Look at digital music services like Pandora or Spotify, which are sitting on rich databases of listeners’ personal music preferences. These services can make eerily accurate predictions of what a listener may enjoy. If you like Leonard Cohen, you may well discover that you like Tom Waits and Nick Cave. Pandora or Spotify can predict that.
There is a fun application at gnod.net which lets you enter a musician’s name, and the site visually serves up artists based on how close users’ preferences are to the artist you chose. It’s like being able to see inside Pandora. Gnod has similar services for movies and literature. Bear in mind that the strength of the prediction is based on the richness of the underlying data. If you happen to like obscure artists, there may not be enough data on the database to make very good predictions. The bigger the data, the better the prediction.
As an artist, if you knew what bands you’re associated with, you could market to these other bands’ fans. If you knew that fans of the other bands purchased their music at a specific store, you could actively promote yourself at that store. If you found their fans also happened to enjoy poetry by Sylvia Plath, you could market your music via her publishers, as well. The universe of Big Data enables the artist to discover new fans as these new fans are trying to discover new music. Very exciting!
4. Does Big Data play other roles in song order on an album?
It does, but probably not in the way you think.
One of the reasons legitimate consumers listen to any tracks at all before buying an album is that they’re willing to pay money for the permanent license to play the music. In a world of ample opportunities for digital piracy, they don’t technically need to do that. They’re just upholding the social contract and abiding by their legal obligations to pay for what they consume.
Unlike legitimate consumers, the dedicated music pirate likely doesn’t care about song order at all before downloading the entire album, and it’s for the same reason that companies are grappling with Big Data. As we’ve discussed in other blog pieces, one of the reasons companies are drowning in Big Data is that storage space prices are collapsing, and the cheaper storage space gets, the more data of less and less value can be economically stored.
Economics is all about allocating scarce resources. When storage space is no longer scarce, there’s no reason not to consume plenty of it. With cheaper and cheaper storage, the pirate should have less and less reason not to grab an entire album, or perhaps even an artist’s entire catalogue.
However, after downloading all that content, the pirate still faces the fundamental problem of Big Data: Now That I Have It, What Do I Do? Do pirates listen to an entire album before deciding that it’s worth listening to? I suppose they might, but I doubt they have the time. Just like everyone else, they’ll find themselves back to square one of deciding whether an album is worth listening to.
If time, not disk space, is their limited resource, pirates will have to find some way of churning through their Big Data quickly. I suppose they could piggyback on Pandora’s analytics to help decide, but I think it’s much more likely that they just do it the old-fashioned way by listening to a couple of songs. Which ones? My guess is that they’ll be the first couple on the album.