Beyond Binge Watching at Netflix

Chief product officer Neil Hunt is dreaming up the next evolution in streaming entertainment at Netflix

A typical American adult trudges home after a long day at work. She’s tired, looking forward to a relaxing evening. Maybe she heats up some leftovers, pours a glass of wine, and plops down onto the couch, opening her laptop, or switching on her smart TV. The red screen appears. Her finger hovers over the queue.

This is the moment of truth, and little does she know that an entire team of data scientists, developers, and designers have spent countless hours dissecting this very moment, building an experience that is somehow individually tailored yet universally appealing. This is the moment that Neil Hunt, the chief product officer of Netflix, knows is crucial to the success of the company, because it’s at this moment that the two sides of the business—the content side and the technology side—are perfectly merged.

“We have a collaboration between a human analytics group and the algorithms and data.”

Whether our American hero will choose to continue watching an episode of Buffy the Vampire Slayer she started yesterday, or try out a new Netflix original documentary about world-renowned chefs, or find an indie flick based on her interest in visually stunning dramas, that moment has been brought to bear by the convergence of content and form, art and technology. It’s been engineered, just for her, and the more than 65 million other Netflix streaming subscribers around the world.

When Hunt talks about bringing the best possible viewing experience to Netflix customers, he’s excited not just by the results his team produces but also by the technology that bolsters them. “The recommendations technology isn’t just a feature; it sits behind absolutely every screen that presents content to the user,” Hunt says. “What are you going to be interested in tonight? What did you watch yesterday, what have you watched in the past that’s like that, where can we draw connections, how can we use the data to put twelve things in front of you, one of which is a really compelling, interesting piece of content that’s going to capture your interest and engage you for the night?”

It takes many smart minds and much trial and error to actually answer that question, but that’s what drives Hunt’s teams at Netflix’s headquarters in Los Gatos, California. As the chief product officer, he’s overseeing several departments that work together on the design, implementation, and operation of Netflix’s streaming technology—figuring out what to build, how to design it, and how to operate it. The teams overlap and collaborate, which is what Hunt says enables them to innovate without fear of breaking anything.

“By keeping the product management and engineering close together, we unlock a lot of synergies, making sure that we design something that’s buildable and build something that meets the real requirements of the design,” Hunt says. “Then on the operational side, if you build the thing and then toss it over the wall to an ops team to run it, you lose the opportunity to really internalize the lessons that come from discovering where and why things break and how they need to be fixed.”

Under Hunt’s guidance, Netflix was an early adopter of “dev-ops,” a system in which the developers of a product also operate that product and are responsible for how it works. The company routinely rolls out new offerings and variations for its streaming customers in small groups, testing them and getting user feedback before implementing changes to the whole subscriber base.

“Developers conceptually carry the pager that wakes them up in the middle of the night if [a component] stops working, which gives them great incentive to build it so that it’s stable; so that if it fails, it fails safely,” Hunt says. “We’ve done a lot of work to build a very available, very reliable system over the years.”

Hunt joined Netflix in 1999, after cofounder and CEO Reed Hastings, with whom he had worked at Pure Software in the nineties, persuaded him to join the still-new company.

36%
Netflix’s share of peak download Internet traffic in North America

65.6M

Number of Netflix streaming members worldwide

$2.6M

Net income in Q2 2015

100M

Hours of content viewed by Netflix members daily

10B

Hours of content watched by Netflix members globally in a three-month period

“It was a radical shift in direction, from enterprise-software engineering to consumer business, and within a year or so, I was actually doing the product management side of the business too, which was a very different thing than I had been doing,” says Hunt, who has a doctorate in Computer Science from the University of Aberdeen, U.K., and a background in research and development. “Back in 1999, there wasn’t really a big population of dot-com experts or people who had worked at this kind of industry for a long time. Everybody who was moving in there was getting in at the ground level and learning new ideas and new technology. That was the norm, and it was really an exciting time.”

Hunt was able to build his department from the ground up, and played a big role in developing Netflix’s game-changing streaming service. Today, Netflix is not only a leader in the delivery of shows and movies; it’s creating new shows, including Emmy-nominated and critically-acclaimed House of Cards and Orange is the New Black, as well as producing fan-favorite reboots of shows once canceled by ratings-dependent traditional networks.

In its 2015 third quarter earnings report, Netflix boasted to shareholders that on the Sunday following the third season release of Orange is the New Black, “members globally watched a record number of hours in a single day, led by Orange, despite the season finale of HBO’s Game of Thrones and game five of the NBA finals also falling on that Sunday.”

While that’s certainly a victory for the company’s programming side, those kinds of record-number statistics also present unique challenges to the engineering and delivery side for which Hunt is accountable. The company that practically invented binge watching is now responsible for 36 percent of peak download traffic in North America, according to bandwidth-management systems vendor Sandvine, creating conflicts with ISPs over who should be paying for this influx of traffic. Netflix’s answer to this was to develop OpenConnect, an appliance that tech news blog Gizmodo reverently called “an unassuming box that holds approximately one Netflix.”

OpenConnect boxes contain large quantities of Netflix’s streaming library—each can be tailored for what’s popular in the geographic region it will serve—and are placed in ISP data centers or Internet exchange locations around the world to deliver content right where it’s most in demand.

“The data doesn’t flow across the Internet backbone, so your video data isn’t competing with everybody else’s email and Facebook,” Hunt explains. “It’s really being delivered over what the network engineers call ‘the last mile’ which may not be an actual mile, but is sort of conceptually the last step of the Internet before you reach the consumer.” OpenConnect helps ensure that thousands of subscribers at once can watch popular shows or movies; it also helps the company to argue in favor of net neutrality, Hunt says, by making it easy for ISPs to pick up the data from Netflix close to the streaming user at no cost.

The company is able to determine what should go in those OpenConnect appliances the same way they can predict what its viewers will watch—through analyzing massive amounts of data. Data analysis is also how Netflix creates what it calls “micro-genres,” the sub-categories tailored so specifically that they can seem spooky. That average American on her couch deciding what to watch may select not only from Popular on Netflix or Comedy categories, but from micro-genres like “dark cerebral foreign crime dramas” or “romantic comedies featuring a strong female lead,” and they’re far from random.

“We have a collaboration between a human analytics group and the algorithms and data. There’s a group of my team, they’re called the taggers, and their job is to watch every piece of content that we have, and then to classify it,” Hunt says. “They have a universe of 500 or 1,000 different tags, like ‘cerebral’ and ‘dark’ and ‘crime drama,’ and they’ll attach those tags, or sometimes they’ll score those tags. How rich is the character development, how visually stunning is this piece of content on a scale of one to five?” Everything gets tagged, and then algorithms do the work of determining how likely a viewer is to watch Show A if they’ve already watched ten other shows tagged with “crime.” Hunt says that a big part of why this works so well is that the Netflix interface uses these tags to tell the viewer why they might be interested in something.

Inclusive Binge Watching

In 2015, Netflix released its original series Marvel’s Daredevil, about a blind lawyer fighting injustice by day and battling criminals as a superhero by night. The most daring part of the series, however, is the fact that it’s accompanied by an option for English audio description. With that option selected, a narrator fills in the gaps for visually-impaired viewers by describing the action in between sound effects and dialogue.

As the show’s first episode begins, the narrator sets the scene: “A man walks cautiously between cars toward an accident, his face searching. Traffic is stopped. Dozens of pedestrians rush to the scene. His eyes grow alarmed as he shoves his way through the crowd.”

The offering is just the first in what Hunt says will be a growing library of films with narrative description. “We’re interested in offering a service that is inclusive for both hard-of-hearing and also visually disabled people,” Hunt says. “We started several years ago doing English captions for content for our American subscribers, and we’ve extended that to include captions and subtitles in up to fifteen languages for different pieces of content across the world.”

The company recently launched in Japan, which Hunt says required his teams to tackle the particular challenges of presenting the Netflix portal and its content and subtitles in a glyph-based language.

With Netflix now currently available in all of North and Latin America, much of Europe, and Australia and New Zealand, and plans to be global by the end of 2016, the company is committed to reaching as many viewers as possible. “We’re working toward having relevant captions for most of our overseas consumers also, in their language,” Hunt says. “That will take a while to get to, but it’s something where we make steady progress year by year.”

“If I just offer you a suggestion and you don’t know what it is, you don’t know whether you’re going to enjoy it or not. But if I tell you that you’re going to enjoy it because it’s cerebral and gritty and a crime drama, then you understand why,” Hunt says. “And if you’re not in the mood for cerebral tonight because it’s been a busy day, then you can choose a different road.”

Anticipating what viewers want, by sifting through data and testing new products, is a big part of what Hunt and his teams do, and that’s not just related to software. Netflix works closely with television manufacturers to build Netflix buttons right into smart-TV remotes. Vendors such as LG, Panasonic, Sony, Toshiba, Hisense, and Insignia that run the latest version of the Netflix app and can meet a certain set of performance criteria—with features like fast app launch and rapid play starts, as well as ease of access to the Netflix app—in turn receive the “Netflix Recommended TV” designation.

“For the consumer, that provides a bright line, that this is a good implementation that’s going to work well,” Hunt says. “Given that there are now about 3,000 different models of TV and Blu-ray player that operate with Netflix, the choice for consumers is difficult.”

If Hunt has anything to do with it, that choice will become even tougher over the next couple of years, as Netflix prepares to offer content for a new generation of viewing technology. Because the company can control the production pipeline for its original content, it can create, for example, films shot to optimize ultra high definition (UHD) and 4K technology, which offers a resolution of nearly 4,000 megapixels.

“The thing that’s unique for Netflix is that we can embrace and support a 4K service long before a cable operator could, or a broadcaster, or a satellite operator, because we deliver individually,” Hunt says. “We don’t have to allocate a new cable channel, or new wireless spectrum, in order to be able to put a new channel out with a new format. And we don’t have to upgrade everybody’s set-top box to receive it. We can do this on a one-person basis.”

Hunt is also looking forward to the next frontier of what’s called High Dynamic Range, a technology that shows brighter whites and darker blacks with more contrast and less banding or contouring in between pixels, making the images seem more vivid. “I’ve seen some HDR examples where the sun shines on a reflective window or a lake or a piece of ocean, and you have to reach for your sunglasses because the reflection is so bright that it really begins to feel very realistic,” he says.

Some of what Hunt envisions hasn’t even been invented yet, but that doesn’t stop him from looking for what’s missing from the viewing experience and working to make sure that Netflix is the company to fill in those gaps. “I’m super excited about what I’m calling ‘surround vision,’ the idea that you can film a show for super-wide field-of-view—perhaps rendered on three screens, one in front of you, one on each side—curving screens spanning 120 or 180 degrees of view,” he says. The main camera and screen would capture the central action, but the surrounding screens would create an immersive experience that puts the viewer into the set. “The horror movies can be that much more scary; the action flicks make you feel like you’re really in there, engaging in the battle.”

Hunt acknowledges the financial and technical hurdles he and his team would face to push that type of technology into the mainstream. But not that many years ago, the idea that a person could sit down at a computer screen, and instantly call up thousands of TV shows and movies to watch at any time of her choosing, also faced significant hurdles. Hunt’s enthusiasm for this next phase of entertainment innovation makes it seem just that much closer, like you could reach out and touch it.

“I don’t know exactly how we’ll get there, but I’m pretty confident that Netflix is in a great position to get on the forefront of that and help to invent some of that technology and begin to bring it to reality,” Hunt says. “I’m excited to be a piece of that, because that will be fun. It’s not today’s mainstream problem, but it’s something that I think is an adventure for the future.”