Thursday, October 23, 2014

Rocktober: Naive Disillusionment & the Data Mining of Music

This summer I attended a semi-exclusive A&R event in midtown Nashville, where a collection of up-and-comers and hopefuls on the country music scene performed for a small crowd and hobnobbed with label reps, hoping to get the right one’s attention and be offered a contract and a path to the big time.

While there, I got into a fascinating conversation with a guy who has begun a start-up company specializing in data-mining hit music.

Basically, he said, making chart-topping music is more science than art. It’s no different than cooking. You need to know the precise ingredients and measurements required to make any specific dish. Then, and only then, can the supremely gifted chefs add their “little something special” that vaults a simple recipe into another stratosphere. But even the “little something special” things can be measured, tracked, investigated, analyzed.

Likewise with music. Hit songs are, truly, recipe driven, and it’s more true in contemporary, mainstream country music than anywhere.

So his company takes hit songs and breaks each one down into hundreds of different data points. Beats per minute. Primary chord and chord progression. Number of words per verse, per chorus, per song. Time length of verses, of chorus, of bridge. If it can be isolated or categorized, they’re doing it. Song by song, note by note, data point by data point, all going into columns and rows for comparison and contrast.

“Once we’ve built up a sufficient database of hit songs, we will be able to take someone’s new song, run it through the analysis, and identify the precise likelihood of that song becoming a hit,” he said. Even now, he explained, with the limited database they’d built up, their predictions were stunningly accurate and were only going to improve as they tweaked the system. They could isolate the comparison to the past year's worth of hits or the past two decades' worth.

Surely they’re not the first ones to try this, to come up with scientific formulas as a means of dissecting hit music, I said. No, he said, but the music industry is not unlike professional baseball.

“Have you seen or read Moneyball?” he asked. And he had me. Because I frappin’ love Moneyball.

“How long did it take baseball, a sport swimming in statistics, whose popularity is built around statistics, to wake up and realize that you could use those same stats to build championship teams rather than relying on the judgment of scouts and GMs? It took a long time, right? Because they were in denial.

“That’s where the music industry is. They’re just waking up to the reality that they’ve spent lots of time crunching music stats, but they’ve been looking at the wrong numbers in the wrong ways and relying on A&R reps and talent spotters rather than using the stats to build championship teams.”

It's cool. It's believable. It's utterly depressing.

Music isn't baseball, dammit.

Inside the scaffolds and structure of a song, the innermost part of myself can seek sanctuary, that endless hunger for inspiration and insight can often be fed. Music is my real church. Like love and faith, I perceive my relationship with music as a mystical experience, not a scientific one. I don’t need to understand why I like a song or a band, only that it moves me somehow in a way I seek to be moved.

My heart wants to believe you can't do this. You can't entirely make art into a science. But I'm sad because, and my heart hurts because it's probably true.

But. If it were completely true, if science was vehicle by which we could all be pulled back together into a shared love of particular songs, why are we more splintered musically than ever? Why is everyone fighting for ever-smaller slices of an ever-smaller pie?

No no. There's still plenty of magic in music. Or, as Olivia Newton-John would say so well, "I have to believe we are magic."

2 comments:

troutking said...

That is depressing, and an argument against a certain someone's admonition that I need to listen to more new music.

Bob said...

Can you get those parameters? I'm trying to write a hit song.