Find Out How Netflix Creates Those Oddly Specific Recommendations for You to Watch

Find Out How Netflix Creates Those Oddly Specific Recommendations for You to Watch

Jan 03, 2014

If you spend any time at all on Netflix’s Watch Instantly service, you’ve no doubt noticed that the website loves to offer you viewing suggestions under some super-specific headings. Stuff like “Cerebral Slasher Movies from the 1980s” is the kind of thing you’re likely to see if you logged into my account.

Many have wondered how Netflix’s algorithm comes up with these bizarre categorizations and just how many of them there are. And now, thanks to the work of Alexis Madrigal, we know the answers to these vexing questions.

Madrigal started out wanting to reverse engineer the Netflix system to find out how many different genre descriptors there were. He uncovered that number (76,897, to be exact), but this led him down a data rabbit hole that eventually found him at Netflix HQ, talking to Todd Yellin – the man behind the system.

The long-form piece Madrigal has written about his experience is sure to be fascinating to movie lovers, computer geeks and anyone who loves huge amounts of data in equal measure. Yellin reveals that Netflix employs a group of people who do nothing but watch movies and give them various meta tags, for example (where do I sign up for this job?), and then gets into the nuts and bolts of what the system is trying to achieve.

Madrigal’s article is filled with some interesting data. Did you know that for some strange reason, Netflix’s favorite actor is Raymond Burr? The Perry Mason star has 19 “genres” devoted him – while Bruce Willis and George Carlin come in second with 17. How about this – would you have guessed that there are more genre tags for movies for the ’60s, ‘50s, ‘40s and ‘30s than there are from the 1990s?

This is all just the tip of the blade when it comes to the fascinating data discovered while mining Netflix’s system. The real beauty of Yellin’s algorithm, as Madrigal points out, is that it not only tells you what you’ll probably enjoy viewing, but gives Netflix a really good idea of what America as a whole likes. This is pretty clever – it allows the company to retain customers by making them feel like Netflix “gets” them, and at the same time, it gives Netflix an inside edge when it comes to developing its own series because they know what people like. That’s pretty incredible when you factor in that there are over 40 million subscribers feeding new info into the system on a daily basis.

Check out Madrigal’s full article for a fascinating breakdown of how it all works, plus a random genre generator that could help aspiring screenwriters come up with a great premise. 




Categories: News, Buzz Bin, Geek
blog comments powered by Disqus

Facebook on