A FORMIDABLE CHALLENGE October 2006, Netflix launched a formidable challenge: whoever would come up with a software that was 10% better than Cinematch, the Netflix software for predicting the movies customers would like, would earn a $1 million prize!
Tens of thousands of people from all over the world started working on this task right away, and each day teams submitted their updated solutions to the Netflix Prize Web page. Netflix instantly calculated how much better than Cinematch they were. Even if there were so many people working on the challenge, it took almost three years before a team came up with results that surpassed the 10% barriers. The winner was BellKor’s Pragmatic Chaos, a seven-person team of statisticians, machine-learning experts and computer engineers from the United States, Austria, Canada and Israel.
THE IMPORTANCE OF THE RESULTS The Netflix Cyberflix tv apk contest was largely followed not just because of the $1 million prize, but because of its lessons which could extend well beyond improving movie recommendations. The data base made available by Netflix for this competition -over 100 million ratings that more than 480,000 users gave to nearly 18,000 movies- was one of the largest real-life data sets available for research. The outcome of this project is most valuable, as a large-scale predictive model can be applied across the fields of science, commerce and politics.
APPLICATIONS: CROWDSOURCING -A MODEL WITH GREAT POTENTIAL The Netflix contest is another project that proves how effective crowdsourcing can be. Before launching this challenge that was open to everyone out there in the crowd, Netflix’s founders had tried for years to improve their software, with only incremental results. So, in this case, the most efficient way of solving the task was neither hiring a highly qualified professional to do the job, nor outsourcing the task (employ a third party to solve the task), but crowdsourcing it.
The progression of the results that were achieved and the way teams came together, especially in the last part of the contest, suggest that this kind of Internet-enabled approach, known as crowdsourcing, can be applied to complex business and scientific challenges.
The blending of different statistical and machine-learning techniques “only works well if you combine models that approach the problem differently,” said Chris Volinsky, a scientist at AT&T Research and a leader of the Bellkor team. “That’s why collaboration has been so effective, because different people approach problems differently.”
Another detail that is very relevant to the power of crowdsourcing is that among the top teams there were not only academic researchers, but also laymen with no prior exposure to collaborative filtering, who were virtually learning the problem space from scratch.
CROWDSOURCING BENEFITS So, we can see that the benefits of crowdsourcing that I enlisted in my previous post apply in this case too. I said there that one benefit of crowdsourcing for the organization is that it can tap a wide range of talent that might not be present in its own organization. And, on the consumer’s side, the benefit is that crowdsourcing opens the door for virtual unknowns in their fields to gain large-scale recognition for their talents. And that is exactly what happened in this case too.