Crafting a Hit Record

Hosted by

This is Celia Hirschman with On the Beat for KCRW.

For decades, the record industry has been in the business of making hit records. It all starts with the executive who has the vision and talent to recognize a star in the making. Then, hundreds of thousands of dollars are spent recording to get the sound just right. Once an album is finished, remixers are often called upon to shift the tempo and create new beats, in the hopes that one of these renditions will hit big with the public.

When the music is finished, the marketing machine kicks into gear with a team of specialists. First artwork, then manufacturing. The online bloggers begin dropping messages about the artist on the net.The publicity officers pitch for coverage. Promotion heads send their troops into battle, to try and convince program directors why the song should be on the radio. And dozens of salespeople are engaged to sell the music into record stores.

This entire process is enormously costly and there are no guarantees that the song will be a home run, no matter how much money, or how spin is applied. The record industry has always regarded the hit making business as a highly risky, unpredictable proposition.

So it's particularly interesting that a company called Platinum Blue claims they can pick hit records in advance. Hit prediction is a very sensitive issue for music professionals. Everyone thinks they have the ears to find the next Beatles or Nirvana, but experience shows that few actually do. Case in point: both bands were rejected by numerous major labels before getting a recording contract.

Platinum Blue does not rely on superior vision or talent to determine hit records, but instead, complicated mathematical algorithums. Using a method they call "spectral deconvolution," the Platinum Blue software can read a song, and within seconds reveal 40 pieces of information about its structure, it's fullness of sound, the instruments it uses, the melody, harmony, pitch, chord progression, cadence, frequency, rhythm, and other inherent qualities. Once they created the software, Platinum Blue began inputing all of Billboard's Top 100 Singles, going back as far as the early 60's. Each song was plotted on a three-dimensional grid, using the mathematical results.

The company then limited data to only the hit records. That's when they discovered the clusters of commonality. Music with entirely different sounds began to register as similar, when considering their mathematical musical equations. It turns out, 80% of all hit songs share a relatively small number of underlying structures, regardless of how different the songs sound. Norah Jones' recent hit belongs in the same cluster as Van Halen's hit. In fact, according to Platinum Blue, contemporary songs by U2 come up in the same cluster as Beethoven.

Which means, hit music is in fact predictable. That should be great news for labels, who can save millions of dollars by testing their songs first before engaging their expensive marketing teams. But record labels are stubborn. Most executives believe they know better than something as mundane as a math equation.

The truth is, hit songs are probably predictable. They feel warm and sounds fresh, yet familiar. That's what makes them popular.

And popular music is not necessarily bad. There's enough room for an artist like the Beatles to write Love Me Do and also pen While My Guitar Gentle Weeps.

Thank God for that.

This is Celia Hirschman with On the Beat for KCRW.