Rethinking Markets

I’m always struck by the extent to which marketers have internalized the assumption that my preferences are stable, and static. Discussing next-generation mobile phone software in Technology Review, Kate Greene writes:

The software, called Magitti, uses a combination of cues–including the time of day, a person’s location, her past behaviors, and even her text messages–to infer her interests. It then shows a helpful list of suggestions, including concerts, movies, bookstores, and restaurants.

The idea is that as the software learns more about you, say, whether you like expensive food or fast food, and it’s 12:30pm, it will use GPS to suggest local fast food joints. The key for this is to predict future behavior based on past behavior.

For businesses, the future is all about personalization, effectively the capture and crystallization of a potential consumer’s preferences so as to sell them more stuff. In a large (retail and business-oriented) financial services firm I know of, the big innovation over the last couple years was personalization. There are quite a few organizations out there that engage in this kind of sell-by-stable-preference-assumption: Poindexter (now X+1), Epiphany, Revenue Science, Tacoda, DART (DoubleClick)

I want to unpack this notion of recommendations, since it is widely held to be the next generation of all sorts of business applications. And the idea of preferences is folded into assumptions about the micro-foundations of all kinds of markets as well. Assumptions about preferences are that they are relatively stable, determined a priori, and randomly distributed.

The idea and main assumption in practice is that by ‘knowing’ the user, and providing them with goods/services that are what they want, orgs can increase sales. This is an empirical question, however, that people want stuff that makes them ‘more’ of what they already are – that is, that these systems try to reinforce existing preferences, while expanding around the edges. Why listen to random radio stations on your cross-country trip when you can listen to your favorites on Sirius from Maine to Montana?

There are other models, of course. Personally, I want new things that are orthogonal to my current stuff. Theoretically, this suggests that people may have preferences for new and unusual stuff. If I like Chinese food today, I will decidedly not want Chinese food tomorrow – but my mobile phone is sure as hell going to recommend a Chinese restaurant. Is there a way to build in future preferences based on heterogeneity rather than homogeneity of past choices?

Opening up multiple vectors of preferences (when I go book shopping, sometimes I want more economic sociology, sometimes I want the most popular, sometimes I want to follow a trusted person’s advice, and sometimes I want to read what Ann Swidler is reading) seems a really fruitful but highly under-theorized direction.

Lena and I, in a fit of greasy Big Nick’s lunch inspiration, had kind of codified this into four distinct paradigms (and I’m cribbing from her characterizations now):

• Self-reinforcing taste: this is what personalization is currently about, that tells us what other purchases our current purchases (or our class, geographic location, race, gender) predict. It is a mechanism that keeps us self-reproducing our tastes “exactly”.
• Popular taste: this is taste driven by fads, and it is what makes us want to be a part of seeing the ‘number one movie in the country,’ or the NYT best-selling books. Assumedly, it also encapsulates quality, Wisdom of Crowds style. But we want to be able to discuss stuff with others around the water cooler.
• Curatorial taste: this is the kind of taste driven by expert discourses, like Ebert & Roper. On the web, I love Kottke for his interesting links. I’m not really like him, and don’t really want to be, but he points me in interesting directions. Some people love BoingBoing for the same reason.
• Aspirational taste: this is why mavens or cool hunters do the do – they help us to see what we need to like in order to pass (or see ourselves as passing) as a Brooklyn hipster, or as an old fashioned academic.

It seems to me that there are empirical questions about which of these models actually reflects peoples’ behavioral and self-conceptions. The stakes are quite high, if you believe the Paradox of Choice arguments, and as choices become more abundant. There is obviously a fascinating and wide wide world out there for ‘recommenders’, including Facebook’s recent recommendation-via-networks (the self-important CEO argues that “Nothing influences a person more than a recommendation from a trusted friend”); A colleague in the wine world ponders recent attempts to adjudicate between varied wine recommendation systems; Netflix offers up a cool $1M for improving movie recommendations based on their current movie preferences; and there is a whole conference/world about recommendations.

This is fertile, fertile ground.