Trust busters and monkey boxing: why Netflix watches everything you watch
Todd Yellin talks about how his company strives to individualise content in order to give us what we want
Screening process: the better Netflix can get to know you, the more likely you are to continue subscribing. Photograph: Gareth Cattermole/Getty Images
Depending on which Jungian theorist you talk to, there are between four and seven basic plots in fiction. For example, Christopher Booker’s Seven Basic Plots lists “the quest” and “rags to riches” but not “drunk-driving, monkey boxing”. The latter is a category, I gather from listening to product development chief Todd Yellin, that could well appear on Netflix’s list of more than 75,000 “microgenres”.
“Does it have car chases? Does it take place in a bar? Is there a monkey? Is there any element of sports or boxing?” the former documentary-maker asks gleefully. He is giving me examples of the kinds of questions Netflix taggers have to ask when tagging a film or television show.
Each movie or TV programme on Netflix is minutely tagged and then appropriately matched to specific Netflix users by a complex series of algorithms. The better Netflix can get to know you, the more likely you are to continue subscribing. So if, for example, you have watched a load of boxing movies with car chases and barroom-based monkeys, they might recommend Thumpy Drunky Monkey Car Crash II (this may not be a real movie) and you would be thrilled. But if they recommended Sober Pacifist Pony Party IV (this may not be a real movie either), you would be disappointed. Yellin calls inappropriate recommendations, which we have all experienced, “trust busters”.
Tagging: the dream job?
Yellin’s job is “to improve the Netflix member experience”, to which end he wrote a document called The Netflix Quantum Theory in 2008 (“we were into pretentious naming”), devised a complex categorisation system and is now recruiting British and Irish “taggers”, a special breed of film and TV buff whose job involves watching hours of TV and film and then analysing the content. It is being proclaimed around the internet as the best job ever.
“Our algorithms work in association with the people who are doing the tagging,” says Yellin. “[Taggers] analyse the heck out of a movie or TV show, and the next step after that is, the algorithm finds what tends to work for individuals. Is the title about business or celebrities or chases, competitions, con games, conspiracies, family life, makeovers, reincarnation or sharks? The movie could be about makeovers and sharks. Then we’ll get down to stuff like: is it visually striking? Is it critically acclaimed? Has it won awards? Is it experimental? Is it independent or more mainstream?”
Certain attributes are rated from one to five. This results in odd moral calculations. “How much gore is in a title? Five might be spleens being ripped out of someone’s guts, while zero is nothing at all gory, perfect for a small child. Those scalar values we’d also use for how ‘cerebral’ a title is or how ‘sad’ or ‘upbeat’ a title is.”
According to a recent Wired article, Netflix’s many categories include “cult evil kid horror movies”, “critically acclaimed emotional underdog movies” and “visually striking foreign nostalgic dramas”.
Tagging Netflix users
Yellin is a little more circumspect when talking about how users themselves are tagged, but says that each has a “fingerprint” based on what they’ve watched before. He has found that people are ultimately unpredictable.
“We’ve found over the years that gender and age are so inconsequential,” he says. Indeed it’s often hard to tell a 60-year-old woman apart from a teenage boy, he says. “This stuff totally transcends gender and stereotype, and we’ve proven that at Netflix beyond doubt. We try not to pigeonhole our users and stereotype them and label them too specifically because that can be confining. There are so many bizarre combinations in terms of what titles [people] watch that nothing surprises me any more.”
Taggers with local knowledge are important. “We really want to get a lot of perspectives,” he says. “So yeah, we’ll get people who are horror aficionados or real experts at romantic dramas or documentaries, but there are cultural sensitivities that are also important to us and we want to capture that. One of our tags is ‘witty’, and when it comes to ‘witty’ you have to have really incisive humour in your dialogue when dealing with the Irish or the British. That’s different when you’re dealing with someone from the US.”
I mention the recent controversy over Facebook engaging in psychological experiments with its users. He stresses that no one else has access to the data Netflix collects.
“We’re very upfront,” he says. “While we develop sophisticated profiles around someone’s taste, the beautiful thing about working for a subscription company is that all the data we collect on someone we only use to make their particular experience better, so they stick to the subscription. Being in the entertainment business as opposed to advertising like [Facebook] allays a lot of those concerns.”
Do some people not find this monitoring of their viewing habits a little creepy? Nowadays, he says, people simply expect internet technology to penetrate every aspect of their lives and they want to receive increasingly personalised content.
“Back in 2006 I used to hear a lot of things along the lines of ‘Big Brother’ and ‘that’s creepy’ and ‘Oh my God, I can’t believe you know that.’ But society evolves.”