Predicting What Your Consumer Wants

The uncertainty principle of quantum physics states that you cannot know the location of a particle and it’s momentum at the same time.

The phenomenon is not a failure of instrumentation; just the nature of the system.

Of a similar vein, I posit the uncertainty principle of digital marketing: you can’t know who a consumer is and their tastes at the same time.  After all, even if consumers would submit to a lengthy questionnaire (which they won’t) what sort of questions would you ask?  I can’t explain my taste for Apple Jacks – the things don’t even taste like apples!

Assessing the value

Knowing what a consumer wants is extremely valuable.  Even more valuable, is having consumers that trust you to tell them what they want. For this reason, Netflix is famously offering up a million dollars to anyone who can improve the accuracy of their Cinematch recommendation engine by 10%.  Sounds like a lot of money for little gain, but most experts agree that the prize money represents a bargain for the company.  The chief executive, Reed Hastings, points out that “getting to 10 percent would certainly be worth well in excess of $1 million.”  Why?  The New York Times describes the problem:

Customers pay a flat monthly rate, generally $16.99 (although cheaper plans are available), to check out as many movies as they want. The problem with this business model is that new members often have a couple of dozen movies in mind that they want to see, but after that they’re not sure what to check out next, and their requests slow. And a customer paying $17 a month for only one movie every month or two is at risk of canceling his subscription; the plan makes financial sense, from a user’s point of view, only if you rent a lot of movies.

Predicting Consumer’s Tastes

The fiscal desirability of a solution is obvious, but how can we predict a consumer’s taste?  Several methodologies have been employed so far, the most prevalent being “collaborative filtering.” By aggregating people’s choices, guesses can be based on correlation.  People who like X also tend to like Y.  So to all the X lovers out there, may I suggest a nice helping of Y?

This works fairly well, but tends to fall apart around love-it-or-hate-it polarizing decisions like whether or not Michael Moore’s movies are masterworks of political dissent or hopelessly biased drivel from the radical left wing of America.  (Whatever your opinion, at least we can find common ground in Tom Hanks films.  America loves Tom Hanks.)

So what would be a useful predictor for those situations?  Research shows that people are more likely to prefer things their social circle [login required] is also interested in.  In other words, to predict your tastes, I don’t need to know what you like if I know what your friends like.  And for that, all I need to know is who your friends are.

Know thy consumer’s friends, know thy consumer

Enter the Facebook Connect and MySpace Data Availability platforms.  If users access sites through either technology, businesses could track both their and their friend’s activities.  It stands to reason that any given set of “friends” will likely inhabit the similar social circles and, odds are, that some of their tastes will be the same.  As such, choices on a site could be given additional prominence if a user’s peers had bought or viewed particular items.  As Netflix understands, merely improving the guesses slightly pays off dramatically.

As with any use of personal data, privacy is a concern.  However, revealing the data is purely voluntary and isn’t anything more than what the user already revealed through the social networking sites themselves.  Across the web, users have repeatedly demonstrated that they are willing to reveal information about themselves for a more personalized and social experience.

Maybe soon we’ll be able to blame our friends for all our online buys.