War is 90 percent information.

—Napoleon Bonaparte

In April 2013, thousands of doctors, scientists, and researchers from dozens of international organizations did what was once considered a pipedream. Led by the National Human Genome Research Institute and the U.S. Department of Energy, this group of luminaries successfully completed the Human Genome Project.

The potential benefits of this accomplishment are manifold and hard to overstate. Perhaps most important, new testing and related data should enable far better predictions of individual health risks, superior means of alternative medicine, and personalized medicine. The possibilities are endless. As it happens, though, the process of comprehensive mapping of complex attributes at the smallest possible level—and critically, the relationships among these attributes—is not restricted to health care.

THE GENES OF MUSIC

In 1999, Will Glaser and Tim Westergren were thinking about a different kind of radio service. What if they could map the “music genome”? If they pulled it off, they could provide remarkably personalized music recommendations. Along with Jon Kraft, the group founded Savage Beast Technologies, the predecessor to the music-streaming service Pandora (launched in January 2000).

At a high level today, there are two ways to provide music over the Internet. Spotify, TIDAL, Google Play Music, and Apple Music all allow their users to select the songs they want to hear. The now-defunct Rdio did the same.^* If you want to listen to a single artist, album, genre, or song all day long, then try these music services. Their core assumption is that listeners know exactly what they are in the mood to hear. Why get in their way? Waiting for random songs harkens back to radio’s early days; instead, customers should serve as their own DJs.

From the onset, Pandora operated under a fundamentally different premise from many of the other legal on-demand music services yet to come. Pandora’s folks saw things much differently. Listeners certainly know the albums, songs, artists, and even genres they generally like, but what if they only like what they know? Not all unknowns are created equal. To paraphrase Donald Rumsfeld, there are known unknowns and unknown unknowns. There are things we don’t know we don’t know. What if Pandora could learn its customers’ listening preferences? What if it could identify interesting music based on detailed listener and song data?

From Theory to Practice

Pandora recognized that one size doesn’t fit all, but it needed to do several things to accomplish its ambitious goals. First, to be appreciably different from terrestrial and even nascent Internet radio stations such as AOL Radio, Pandora needed to quickly make personalized recommendations—and good ones. Second, it could not rely on simple factors such as a song’s genre and the year that it was released. To this end, Westergren and company worked on what it dubbed, and eventually trademarked, as the Music Genome Project.

According to its website, Pandora believes that:

. . . each individual has a unique relationship with music—no one has tastes that are exactly the same. So delivering a great experience to every listener requires a broad and deep understanding of music.

Our team of trained musicologists has been listening to music across all genres and decades, including emerging artists and new releases, studying and collecting musical details on every track—450 musical attributes altogether.^*

Each song’s “gene” corresponds to a very specific characteristic. Examples include:

• The gender of the lead vocalist
• The prevalent use of groove
• The level of distortion on the electric guitar
• The type of background vocals^†

Interestingly, Pandora maps different genres—and different numbers of genes—to different music genres. For instance, with more than 400 genes, jazz is far more complicated than rock and pop (150 each). As Rush guitarist Alex Lifeson once famously riffed during a show, “Jazz is . . . weird.”

Equipped with this vast trove of data, Pandora could then build its product to reflect its core belief. The service made it easy for users to create their own radio stations by selecting a single artist. The service then dutifully serves up its best guesses—that is, songs that its algorithm believes matches the current artist. This serves as only the starting point. As Figure 14.1 shows, users can vote songs up or down, and skip the recommendation.

Pandora is intent on learning as much about your music tastes as possible. Pandora’s founders correctly believed that even the most hardcore audiophiles couldn’t possibly know the name of every song, band, and genre that they would like. There’s plenty of vast ocean of undiscovered music out there—if people only knew about it. As the following story illustrates, I, for one, am eternally grateful for one of its recommendations.

THE TENSION BETWEEN DATA AND DESIGN

There’s always been more than a bit of tension between data folks and their design brethren.^* As companies such as Amazon, Facebook, Google, and Yahoo! have discovered, it’s not easy to strike the right balance between thoughtful human curation and cold, data-driven algorithms. Facebook learned this lesson with its NewsFeed. (See “Automation: Still the Exception That Proves the Rule” in the Introduction.)

Douglas Bowman is “Internet famous” because of how he reacted to this very tension. The Google designer abruptly quit the search giant in 2009, frustrated with, in his opinion, the company’s unhealthy obsession with data. In his words:

Yes, it’s true that a team at Google couldn’t decide between two blues, so they’re testing 41 shades between each blue to see which one performs better. I had a recent debate over whether a border should be 3, 4 or 5 pixels wide, and was asked to prove my case. I can’t operate in an environment like that. I’ve grown tired of debating such minuscule design decisions. There are more exciting design problems in this world to tackle.¹

Since its inception, Google hasn’t been the ideal place for design purists—the anti-Apple, if you will. After all, the company has always been all about data. Up until recently, it did not sell any tangible products. Without perhaps the world’s largest trove of information, Google might not even exist. You can make the same claim about Facebook. These companies live and die by their data. That’s one way to run a business—and today, it’s hard to argue with the results of such a data-oriented approach. Of course, not every organization operates under this belief.

Exhibit A: Apple explicitly rejects data-based design. Steve Jobs believed that it was “really hard to design products by focus groups. A lot of times, people don’t know what they want until you show it to them.”² Jobs wasn’t alone in his unwavering belief in the primary importance of good design, data be damned. Tim Cook has largely continued his iconic predecessor’s design-centric vision. The new head honcho has eschewed using customer data in any way, even when threatened with court orders after the San Bernardino terrorist attacks. (See “The Primacy of Privacy” in Chapter 1.)

NEXT

This book concludes with a look forward. We may not know exactly what’s coming, but that doesn’t mean that we can’t prepare. To paraphrase the iconic Rush song “Tom Sawyer,” even if changes are not permanent, change is.