Netflix is a case in point: In 2006, the movie-rental company wanted to improve its recommendation algorithm, but internal efforts had plateaued. So they took a previously proprietary data set of a hundred million anonymized user movie ratings and published it, while announcing that the first person or team who could use that data to beat the current algorithm’s accuracy by at least 10 percent would win a $1-million prize. Even the contest was open: Netflix reported top teams’ progress on a public leaderboard, and within three years a winning solution emerged.
