I don’t intend for my weblog to become a techno/trend rag, but I’ll admit to a strange fascination for how new technologies change our behavior, or expose behaviors that have always been there. One particularly interesting question for me is how we cluster our cultural preferences (and how arts organizations do it for us).
The traditional performing arts center season, for example, is often broken down into ‘tracks’ or smaller subscription series, usually clustered by genre (Broadway series, chamber music series, dance series, world music series, etc.). Programmers figure that if you like The Producers, you’ll like Thoroughly Modern Millie, and if you like Youssou N’Dour, you’ll like the Lincoln Center Afro-Latin Jazz Orchestra.
These genre categories have always seemed a bit forced to me, and a little too convenient for the presenter/producer. These are clusters on the supplier’s terms, not the audience’s terms. What if I like some Broadway, but also a certain Latin performer, and the Bulgarian Women’s Chorus, and an obscure folk performer of the 1970s, and an angular jazz pianist…tell me which subscription series is best for me (and don’t just say ‘pick five’…assume I’d like a little help).
On-line companies like Amazon have been taking another tack on this question for quite some time now. By tracking your actual purchases, and comparing your patterns to the millions of other patterns they’ve seen, they can recommend a cluster of books that might be up your alley (not based on some predetermined genre, but on the patterns of actual preferences).
These companies use a technique called collaborative filtering (see a definition here, or a previous weblog entry about the technology) to compare your purchase and preference patterns to other users, and to suggest items they also bought. I’m still looking for a performing arts organization that has tried this technology using its ticket purchase data, to see what seasons emerged from actual purchases. If you know of one, let me know.
The new wrinkle on collaborative filtering is that it’s no longer just the domain of ‘big brother’ retailers, and is becoming a tool for personal discovery. Services like Audioscrobbler will watch what music you choose to play on your computer, and suggest other music you might enjoy. There’s no sales engine lurking to sell you something. And, in theory, there are no hidden contracts or commissions leading the software to nudge you toward certain works.
Some will find this self-installed voyeur disturbing. Others will be eager to discover music that matches their ‘pattern’ that they would never find on their own. Either way, it’s another technology to watch. And it’s another reason to question your assumptions when you’re clustering next year’s season brochure (like, how about the ‘Country Music/Grand Opera/Taiko Drumming’ series?).
This concept reminds me of something I read in Trend Watching; they label it as twin-sumer. We so value input from others (either via blogs or person-to-person, but so much more info is available from blogs now it’s a major influence) to help guide our buying and decision-making. Even small companies have hooked on to “If you liked -blank- then you’ll like -blank-” concept. But you’re correct, if your tastes vary, tracking a consumer’s next potential purchase will throw computer logic out the window.
As a programmer, I’ve always been a fan of the choose-your-own series; but the reason hit home for me when we put the ’96-7 season on sale in Wausau. I spent a lot of time that year looking at incoming series order forms and found, to my astonishment, a substantial number of people who paired “Singin’ in the Rain” not with “Ain’t Misbehavin'” (it was the year of the apostrophe), “The Odd Couple” or Hal Holbrook’s “Mark Twain Tonight!” but with “The Song and Dance Ensemble of Tibet from The Snow Fields of China.” It was at that point that I realized that you REALLY cannot anticipate somebody else’s taste.
The difference, however, between Amazon or iTunes tracking your taste in books or music and a performance series tracking your interest in shows is the available-date factor: To fulfill a minimum purchase requirement, a significant number of patrons will choose a 4th or 7th show that fits their calendar, not necessarily their taste. So the record of a regular attender is strewn (like the Snow Fields of Tibet) with anomalies.
That said, I’m intrigued by the possibility of a past sales analysis tool more sophisticated than “If you like peanuts, you’ll like Skippy.” Thanks for the heads-up.