Asking for a friend ... is it convenient or pathetic when Netflix knows your preferences better than your spouse? Or when the only product recommended by Amazon is cat food?
It is no secret that these large e-commerce platforms have an extensive amount of individual user data to draw upon for predictive modeling. That’s incredibly valuable. But is it the only way to predict preference?
As the scrappy, bound-pushing, quantitative researchers we like to think we are — we set out to find a way to predict preferences for more than 4,000 items while knowing only a small handful of actual preferences.
In a representative survey of 2,000 Americans, each respondent was asked if they like, dislike, or are unaware of 50 random items from a list of thousands of books, brands, TV shows, movies, and music artists. With only that information, we developed our proprietary “Bullseye” prediction model. Let’s see how far off the mark we are.
Ignoring predictions for a moment, general preferences pass a sniff test:
- Top Preferences: Jaws (the movie), Cher, Chevrolet, 7-up (the beverage), Elvis Presley, The X Files, Alicia Keys, and woodworking (the hobby) are all classic Americana.
- Bottom Preferences: Royal Caribbean (for obvious reasons), The Man in the High Castle, stamp collecting (again, the hobby), and MIKA (“Relax, take it easyyy”) likely aren’t dinner table conversation starters.
Our initial predictions between brand preference and media preference reveal that Americans who like:
- Coca-Cola also like Braveheart, Breaking Bad, and Blood Diamond.
- McDonald’s also like No Country For Old Men, Modern Family, and Pearl Harbor.
- Budweiser also like Catch Me if You Can, Blade Runner, and Butch Cassidy And The Sundance Kid.
While we’ve got a lot more polishing to get “Bullseye” ready for primetime viewing, you can expect regular updates and insights along the way.