Beats Disrespects Pandora, Talks Too Big a Game

NEW YORK (TheStreet) -- As I noted Tuesday at TheStreet Beats Music is a welcome and, most likely, successful entry into the Internet radio space. As a consumer, I'm not sure why you would cop an attitude other than the more, the merrier.

Even competitors should be happy when legitimate players invade the sector.

From both consumer and competitor standpoints, I actually find it more annoying when companies such as Samsung, Nokia (NOK) and Blackberry (BBRY) do Internet radio and it flops. But, for as annoying as that can be, it also validates the notion that white label, part-time efforts simply do not work. You have to be a music-centric technology company first (like Beats, Spotify, Pandora (P)) or Apple (AAPL) to make it work well and/or achieve scale.

In other words, the crowded space and subsequent failures by massive corporations lends credibility to Internet radio's present leaders. It remains to be seen if Beats can join that club.

But the founders are certainly talking a big game. And I can understand why. They're smart guys with proven track records -- Dr. Dre, Trent Reznor, Jimmy Iovine -- and genuine interest in making this thing work for the benefit of musicians and listeners. It's actually pretty cool to see these guys giving it such an aggressive and enthusiastic go.

But that enthusiasm and its attendant tunnel vision leads to these ultra-successful and confident men talking a book that's really a bit misleading. Consider this excerpt from a The New York Times story on Beats Music:

The idea is that bold visual appeal and the expertise of its programmers (or -- curators, -- in its preferred buzzspeak), in serving up just the right song or playlist, will create excitement among the millions of listeners who have been unseduced -- or just confused -- by streaming music ...
To hear Mr. Iovine and his colleagues tell it, the world of online music is a letdown, full of bland websites, robotic recommendation programs and not enough soul in the machine. Pandora, for example, uses automated musicological analysis to decide what songs to play, and Spotify and others recommend music to users by parsing huge pools of data.
Ian Rogers, the chief executive of Beats Music, argued that these systems inevitably fail because they rely too heavily on computer algorithms and because the people behind them just misunderstand music. He cited one typical, so-obvious-it's-wrong recommendation as proof of the problem: Paul Simon fan? Check out Art Garfunkel! ...
But Beats uses algorithms, too, as part of how it customizes the songs it sends users based on their profiles and listening habits. The difference, Beats executives say, is that their service makes greater use of its editors and guest programmers like Rolling Stone, Rap Radar and Pitchfork, and only recommends the good stuff.

Let's go paragraph by paragraph to cut through the aforementioned enthusiasm, which has a way of creating euphoria, which has a way of creating myths that perpetuate without context and proper pushback.

First, "The idea is that bold visual appeal and the expertise of its programmers (or -- curators, -- in its preferred buzzspeak), in serving up just the right song or playlist ..."

I really want to know how Beats intends to do this, at any meaningful scale, without relying predominantly on technology (a.k.a. multiple algorithms). Is Trent Reznor, alongside a team of millions, going to handpick the next "song or playlist" for the millions upon millions of users Beats surely hopes and expects to attract?

No.

They're going to use something that looks a lot like, but isn't anywhere near as far along or sophisticated as the Pandora model. Computer models at Beats will ultimately place users into aggregate groups on the basis of the information these users input into the platform and their subsequent listening history. Like Pandora's Music Genome Project, which was built and is maintained by humans, Beats will input its own human flavor to basically guide its algorithms (I assume they'll use more than one) on how to proceed in subsets of cases from the overall Beats user population.

I know that doesn't sound as sexy as our curators will serve up the best next song, but it's the truth.

And I ask the same question in response to the second paragraph from the excerpt. Exactly how will Beats service millions of listeners without "parsing huge pools of data." The only legitimate answer to that I can think of is that, relative to Pandora and Spotify, Beats does not yet have "huge pools of data" to parse. That's probably the biggest hurdle the company faces. It takes time for personalization and discovery to get good. And, again, it's not possible to execute, at scale, with anything other than humans building, informing and maintaining the platform and algorithms running the larger show.

This is exactly what Pandora does. I know this. And I have about one hour, nineteen minutes worth of video to prove that the Beats characterization of Pandora's Music Genome Project is littered with falsehoods. That Beats, in all reality, will be attempting to do precisely what Pandora has mastered, but from a considerable position of weakness because of lack of time on the job and, subsequently, rich data.

Ian Rogers was cute when he made the comment about Internet radio platforms suggesting Art Garfunkel to Paul Simon listeners (Pandora, for the record, recommends Simon & Garfunkel, but not solo Art Garfunkel, in Paul Simon Radio), but he really should have watched my videos where Pandora's Chief Scientist Eric Bieschke discussed this very issue. He might have learned something!


In the two videos, Bieschke explains how Pandora's platform works.

When you first interact with it, it provides relatively generic recommendations, both at the station and overall level.

So, when you punch in an artist's name -- say Paul Simon -- to create a new station, the Music Genome Project starts the process of getting to know what type of Paul Simon listener you are. Pandora's going to use the feedback you provide to determine where you fit musicologically.

In the videos we use the examples of Taylor Swift and a Swedish band called First Aid Kit. As Pandora collects data from you, it will determine exactly how to craft your station. As we discovered, Pandora puts artists such as Bruce Springsteen and John Mellencamp on my Taylor Swift station whereas Bieschke gets more dance music and other Pandora listeners receive a decidedly more country experience. There's no one-size-fits-all because Pandora is not a taste maker. It serves music on the basis of what you like, not what its employees prefer.  

So then to the fourth paragraph of the excerpt, and in closing, Beats sounds like it wants to be a "taste maker." Again, Pandora is on record -- in the video series I did and elsewhere -- as saying that's not its business. As Bieschke put it (or did I hear this somewhere else?), Pandora does not want to be "the arbiter of cool." Never has been. And, while it will recommend music to you on the basis of your listening history, it probably will never go very hardcore in that direction.

All of this to say what it pains me to keep saying -- enough of the hubris from guys like the ones at Beats. Your service will be different than Pandora's. And it will go down as an exercise in subjectivity as to who thinks it's better or worse. But there's a decent chance Beats Music might not even exist if Pandora hadn't come before it.

Trent Reznor probably wouldn't want to be part of cutting down artists who came before him and Nine Inch Nails. I'm not sure why he feels the need to malign Pandora to build up Beats.

But if that's how you're going to go about it, fine ... just provide some context. Be accurate. Check yourself. Don't mislead. I don't see the guys from Pandora and The Echo Nest (the service that powers many competing Internet radio platforms) running around pounding their chests. In fact, I reckon these guys wish they could work together more. 

As it ends up there might not be a service that uses algorithms that works in more human curation than Pandora does. We have already generally established who built and maintains the Music Genome Project, but we should note -- as I have in the past and Beats executives conveniently failed to do -- that real live music analysts input four to seven pages worth of musicological data for every single song that enters Pandora's catalog. That's a primary reason why the MGP is so good at determining which song to play next.

It's such a painstaking process, that it sometimes takes a while. I spoke to an artist last week who called his distributor to see when his new record would show up on Pandora. The distributor said, "Remember, it's all done by hand so it's in the queue." I received a similar query from a woman whose daughter will soon be a Pandora artist.

They have to be patient. And Beats better be prepared to exercise the same virtue. They're making this personalized radio and music discovery thing sound mighty easy.

It's not.

--Written by Rocco Pendola in Santa Monica, Calif.

Disclosure: TheStreet's editorial policy prohibits staff editors, reporters and analysts from holding positions in any individual stocks. Rocco Pendola is a columnist for TheStreet. Whenever possible, Pendola uses hockey, Springsteen or Southern California references in his work. He lives in Santa Monica.

More from Opinion

It's Dumb to Think Legalizing Weed Is Still a Political Issue

It's Dumb to Think Legalizing Weed Is Still a Political Issue

AAP Exclusive: Cramer Says The President is No Longer on the Side of the Bulls

AAP Exclusive: Cramer Says The President is No Longer on the Side of the Bulls

Why It Makes Perfect Sense for Netflix and Amazon to Buy Up Movie Theaters

Why It Makes Perfect Sense for Netflix and Amazon to Buy Up Movie Theaters

2 More Reasons to Sell All Your Stocks and Run Away

2 More Reasons to Sell All Your Stocks and Run Away

Sean Hannity's Link to Trump Lawyer Raises Questions: Doug Kass Insider

Sean Hannity's Link to Trump Lawyer Raises Questions: Doug Kass Insider