In 2006, when Netflix was just a DVD rental service, it offered a $1 million prize to the team that could improve Netflix’s movie recommendation algorithm by ten percent.1 The competition became an academic lightning rod—thousands of teams entered and their work produced dozens of academic works.2 During the first year of the competition, several teams made significant headway; for example, AT&T’s team, KorBell, improved the algorithm by 8.43%.3 But soon, progress stalled. Teams snowballed in size as they pooled their efforts in attempts to gain a few fractions of a percentage point.4 Three years later, two teams finally passed the 10.00% improvement mark: BellKor’s Pragmatic Chaos, a hybrid team of KorBell and Big Chaos, and The Ensemble, a twenty-three-team super group.5

Continue Reading