The Day The Music Started

Posted by Mike Walsh

12/22/05 3:14 AM

My favourite CD shop when I was a kid was small and selectively stocked. Small handwritten comments adorned every cover, and the owner had a forensic ability to make music recommendations that bordered on precognition. Unfortunately his story ended on a familiar note. Run out of business by a chain store. Now there is the new generation of online recommendation networks chasing the Holy Grail of consumer preferences analysis. Namely – can you help people figure out what they like?

Using technology to profile consumer preferences is nothing new. Amazon pioneered it with books, Netflix with DVDs and TiVo with TV programs. The approaches vary, but are generally based on a statistical method known as collaborative filtering. By comparing your transaction history with customers with similar purchasing habits, a recommendation system can then predict other products you might like. Sometimes that works well – people who enjoyed watching action extravaganzas like Terminator 3, will probably enjoy Mission Impossible. But there are certainly limitations to that approach – identity confusion being a central one (as infamously reported in the Wall Street Journal article “My TiVo Thinks I’m Gay”). Buying books for your wife on your Amazon account or letting someone else use your remote on you TiVo can suddenly make a supposedly clever recommendation service seem more artificial than intelligent.

Apparently – the ipod changes everything. Music suggestion services are running hot again. Pandora, Last.FM and Music Strands are all recently funded and offer varying degrees of music advice. Pandora in particular – has taken the unique tactic of employing music analysts to study the underlying structure of recordings to classify them based on set criteria such as speed, rhythm, and even rapping style. These elements, or DNA, then form the basis for the system to match your music tastes to new music based on your current listening patterns. Enter a song you like, and Pandora will tell you another song you ought to like as well. It’s intriguing if not a touch totalitarian.

While it may be fun daydreaming of magic equations that predict consumer tastes, a more fruitful path might lie in the less scientific, grassroots community networks organised around music interests. Both Last.fm and Music Strands have online friendship and group forum functions built into their service – but there is no better current example of social networking and music at work than MySpace. First and foremost, MySpace is about finding, chatting, and co-creating content with friends. However, somewhere along the line music also became a key part of the service. So, although MySpace is not designed to actively recommend new music, the same result is achieved by connecting with other users and sharing the experience of music and bands with them.

The shared experience of entertainment media is an area that has been overlooked until recently. If you look at the companies that Yahoo has been buying such as Flickr (community photo sharing) and del.icio.us (community content tagging) – a powerful subtext is that people not only want to create their own content, but that they also want to share it with each other. One of the prevailing lunacies from some Web 2.0 crack addicts is that community generated content will overwhelm anything created by a publisher, movie studio or record label. But if you look closely at the content of many popular blogs today, and you will see how much of the discussion centers around popular movies, recording artists or books. In other words – we are witnessing the shared experience of media rather than its demise.

That said – where future recommendation systems will make themselves most useful is in the area of emerging rather than established artists and high budget entertainment products. Everyone has heard of Madonna, Star Wars and Harry Potter, and probably already has a clear view about whether they want to consume those products or not – irrespective of what their friends or an online recommendation service might say. However, what is really interesting and hard to search for are those hot new acts and talents that only a few people have heard of.

Seen in this light, recommendation is actually a subtle problem of search rather than one of taste comparison. The problem is complex because it cannot be solved by spiders indexing content or a set of clever probability algorithms. In an emergent social network based on music, the most effective filters are actually other people. This is not a theoretical argument. You can see this trend in action. The reason that over half a million bands are using MySpace to market themselves today is that it is a highly effective platform for recruiting fans, distributing music and encouraging an ever widening circle of conversations about your brand. It is a total reversal of the usual way of thinking about consumer purchasing behaviour. Don’t search for the music. It will find you.

Topics: Media

New call-to-action

Latest Ideas