The music industry has always been notoriously unpredictable, and the old A&R maxim that the cream always comes out on top is far from a fact. For any band that makes a living from their music, there are at least a thousand that never will, and the proportion of musicians who actually get rich through their work is even smaller. However, there is a general feeling (if not a real consensus) that the musicians who make it are there because, in some way, they are inherently better than the swathes of artists left in their wake.

This calls to mind Robert M. Pirsig’s question about quality: what makes something good? Is there really any objective standard by which that quality can be measured? Most people would say yes, as they can easily tell if a band is awesome or a bunch of untalented hackers, but when it comes down to it, this is just personal taste and opinion. Although certain technical qualities such as musicianship, structural complexity and production values ​​can be pointed to, the music is more than the sum of its parts: the Sex Pistols cannot be dismissed as not having the technical genius of Mozart, as neither can Stockhausen’s music be effectively ranked above or below that of Willie Nelson. It seems that when it comes to music, you have to instill in it a Philosophik Mercury that is as intangible as it is unpredictable. The only barometer by which we can judge is whether we like it or not. Or is there something else?

Recent history is littered with examples of works and artists that are now considered classics (or at least have become wildly popular) that were at first outright rejected by talent scouts, agents, or industry executives. Harry Potter, Star Wars, the Beatles – they all fall into this category, as does Pirsig’s classic work. Zen and the art of motorcycle maintenance, which was rejected 121 times. If phenomena of this magnitude could be overlooked, then what chance do artists of moderate talent ever have of being noticed? On the other hand, the entertainment sphere is teeming with performers who could never hope to be anything close to moderate talent. So does the entertainment industry really know what it’s doing, when so many of its predicted hits fail miserably and rejected unknowns keep popping up with hits on the charts? Recent research seems to suggest not.

Now that Web 2.0 is in full swing, social media is changing the way we access and perceive content. The age of digital music is upon us, and the ease with which new music can be obtained from unsigned bands has created a new economic model for distribution and promotion. Buzz itself is the latest buzz, and word-of-blog/IM/email has become a very powerful tool for aspiring artists. Combined with the fact that individual downloads now count towards a position in the official song chart, the cycle of promotion and distribution of new music can be carried out entirely online. But does such baffled convenience make it easier to predict what will become a hit?

The standard approach of major labels is to emulate what is already successful. At first glance, it seems like a perfectly valid strategy: If you pick a woman who looks a bit like Shania Twain, give her an album of songs that sound the same, an album cover with a similar design, and spend the same amount of money to promote it, then surely this new album will also be a success. Often, however, this is not the case; instead, another woman who possesses all these characteristics (with music of a similar quality) appears out of nowhere and continues to bask in a spell of pop stardom.

This approach is clearly flawed, but what’s the problem? It is this: the assumption that the millions of people who buy a particular album do so independently of one another. This is not how people (in the collective sense) consume music. Music is a social entity, as are the people who listen to it: it helps define social groups, creates a sense of belonging, identity and shared experience. Treating a group of such magnitude as if it were just a compilation of discrete units completely eliminates the social factors involved. While a single person, away from social influences, may choose to listen to Artist A, the same person in real life will meet the artists through their friends, either locally or online, and will end up listening to Artist A instead. Artists C and K, which may be of similar (or even lower) quality, but that’s not the real point. Music can have as much to do with image as with sound.

This raises more questions about quality: Is a song’s popularity based on some kind of chaos theory, all things being equal? Certainly, there is a cumulative advantage effect at work when promoting music: a song that is already popular has a better chance of becoming more popular than a song that has never been heard before. This is clearly seen on social networking sites like Digg and Reddit, where an article’s popularity can grow steadily until it reaches a certain critical mass of votes, at which point its readership suddenly explodes and it goes viral. Such snowballing effects have been known to bring fairly robust servers to their knees with incoming traffic.

Duncan J. Watts and his colleagues recently conducted a fascinating study on the effects of social influence on an individual’s perception and consumption of music. The process was described in an article in the NY Times. Using their own Music Lab website, they studied the behavior of more than 14,000 participants to determine what factors influenced their selections.

Participants were asked to listen, rate and, if they wanted, download songs from bands they had never heard of. Some of the participants only saw the names of the songs and bands, while others also saw how many times the songs had been downloaded by previous participants. This second group, in what we call the social influence condition, was divided into eight parallel worlds so that participants could see previous downloads from people only in their own world. We didn’t manipulate any of these rankings: all artists in all worlds started out identically, with no downloads, but because the different worlds were kept separate, they subsequently evolved independently of one another.

Although the article does not provide information on the demographic details of the sample audience, given the nature of the medium (an online music site that assesses user behavior on online music sites) and the size of the sample, It is probably fair to assume that the results are reasonably indicative. It turns out that the study produced some very interesting revelations:

In all social influence worlds, the most popular songs were much more popular (and the less popular songs were less popular) than in the independent condition. At the same time, however, the particular songs that became hits were different in different worlds, just as cumulative advantage theory would predict. Introducing social influence into human decision-making, in other words, didn’t just make the hits bigger; it also made them more unpredictable.

Consistent with these results, an individual’s independent evaluation of a song is a much less significant factor in its success than social influence factors. The intrinsic quality of a song, if it can be measured at all, is outweighed by cumulative advantage, meaning that a few key votes at an early stage can radically alter the course of the overall selection process. This has some significant implications for musicians, producers, and promoters. Essentially, it means that no amount of market research can allow you to accurately predict which songs will be successful. The behavior of a few randomly chosen individuals early in the process, whose behavior is itself arbitrary in nature, is eventually amplified by cumulative advantage in determining whether a song advances to the next level. The randomness of such a process means that unpredictability is actually inherent in the