OPINION: Are Pandora’s Music Genome Operational Costs Worth It?

Posted December 5

Welcome to the O Music Awards guest writer series, a place where we hand the proverbial reins over to qualified writers/musicians/etc and let them share their thoughts about music, technology and more. Today’s guest blogger is Andrew Richards, from the band Uniform Motion.

Pandora has been in the press a lot recently due to its involvement in what has become known as the Internet Radio Fairness Act.

The purpose of the Act is to lower the statutory rates paid to the performers and master owners of songs played on Internet Radio. These rates were set by the 1998 Digital Millennium Copyright Act.

Pandora believes these rates are unfair because they are higher than the rates paid by satellite and digital radio stations. One cannot refute this. The rates are definitely higher. The question is whether or not they are fair and the U.S. Congress will be the judge of that.

I’m a musician, I own the masters to my songs and I don’t have a strong opinion either way. I would tend to agree with Zoë Keating that I’d rather be paid in data anyway. Bandcamp provides valuable data, but doesn’t pay for the music streamed from the artist’s page (they don’t make any money from it either) and that works our really well for me.

Deezer is just about to launch Deezer for Artists, which is a step in the right direction. I’d love to see Pandora and Spotify do the same.

As for the lowering of rates, I fully understand Pandora’s position. They’re a public company now. Their purpose is to make money for their shareholders. They haven’t been able to generate a profit so they’re looking for ways to cut their costs. Music, the product they’re selling, is costing them 50% of their revenues and no matter how much money they make, that number will not change.

So what can they do? They have a few options in order to please their shareholders. Either they start making a lot more money, they lower the rates they pay for the music, or they lower their other costs. I’m surprised no one has suggested they lower their operational costs yet because this is where I believe Pandora’s business is totally flawed.

I first came across Pandora in the early 2000s when I was interested in finding out more about music discovery tools. Pandora was founded in order to market a playlist generator based on the Music Genome Project.

The Music Genome Project is a fascinating approach to music discovery as it uses a set of musical characteristics to create a sort of genetic fingerprint for a song. Four hundred attributes, such as genre, type of instruments, type of vocals, tempo, etc. are taken into account to form the song’s DNA. Once a song has been analyzed, Pandora can compare it with other songs in their database and create a playlist of similar songs.

What I like about Pandora’s approach is that there is no discrimination in the discovery algorithm, which means that if your song sounds a lot like Coldplay, you have a chance of finding yourself on the Coldplay channel on Pandora. It doesn’t matter if you’ve only sold 12 records. What I dislike is that this approach to music discovery doesn’t push the boundaries much, so you’re pretty much destined to listen to a clone of a clone of a clone.

Last.fm on the other hand, has a different approach to music discovery. It uses the AudioScrobbler plugin to track people’s listening habits. So if you listen to a lot of country music and occasionally listen to classical music, the Last.fm music recommendation system might slip in a few classical songs when a country music fan is after new music.

However, the major difference between the two is that Last.fm’s technology is a software plugin with crowdsourced data and statistics, whereas Pandora’s data costs them a huge amount of money.

Each song has to be analyzed by a musician in a process that takes roughly 20 minutes. As of October 2011, Pandora had 900,000 songs in their database.

The average hourly wage in the US is about $23, so let’s assume the approximate cost is $10/song. That means that Pandora could have spent roughly $9,000,000 analyzing songs so far, but possibly more. iTunes has 20,000,000 songs. It would cost Pandora approximately $200,000,000 to build a library of songs as big as iTunes based on their Music DNA approach.

In my opinion, Pandora’s business has always been flawed because their technology is not really a technology. It’s an expensive man-made database. It seems rather unfair to me that artists should have to suffer a decrease in rates so that Pandora can afford to pay musicians to analyze songs!

If Apple, Amazon or Google move into the Internet Radio space, Pandora is going to be in a very difficult position because all three companies have powerful recommendation engines that do not require a musician to analyze a song for 20 minutes.

This is why I don’t think lowering the rates will save Pandora in the long run.

Image courtesy of Christian Reed