Do Adjust Your Set: Netflix has built 50m TV channels. For founder Reed Hastings, that’s just the start — WIRED

In a production office a few metres from the designer boutiques and tourists on the Rue Royale in Paris, Pascal Breton, a veteran French TV-drama producer, shifts restlessly in his chair as he describes his vision for his latest series, Marseille. An eight-episode Netflix Original series, due in late 2015, Marseille will be the US video-on-demand service’s first foray into French-language production, as it bids to replicate the success of the David Fincher-directed House of Cards.

“When I met Netflix, I said my pitch is very simple,” says Breton, an exuberant television veteran with more than 1,000 hours of French TV to his name. “I said you take House of Cards and Boss [starring Kelsey Grammer], two great US series about power and politics, and relocate them to Marseille, the only city in France with so much potential to just… blow up.”

Marseille is the sort of immersive, multi-layered drama that Netflix, through deep mining of user data, knows its subscribers seek out — and an example of how the technology platform is reinventing screen entertainment. With over 53 million customers in nearly 50 countries, Netflix streams two billion hours of TV and movies a month and is reshaping the way content is created, consumed, delivered and paid for.

Since software entrepreneurs Reed Hastings and Marc Randolph founded the company in California in 1997 as a DVD rentals and sales website, Netflix has transformed from insurgent to behemoth, with a valuation at the time of writing of around $28 billion (£18bn). In the process the founders have eased the stranglehold of legacy broadcast companies — including network, cable and satellite players — on consumers and their viewing habits.

As Hastings tells it, Netflix’s breakout moment came 13 years after the company was founded. In 2010 the company had just launched into Canada, its first new territory. Until then, the Netflix model had been a hybrid of DVDs — which were ordered online and delivered via the post — and streaming. But north of the border they opted, with some trepidation, to go to market offering a streaming-only service.

“It was a beautiful day in Toronto,” Hastings says, sitting in the Netflix office in Amsterdam, “and because we’d spent the day doing demos, we didn’t know what our [Canadian launch] numbers were. That night we got them and found that the number of sign-ups was like ten times larger than we’d thought it would be. That was just shocking and I remember thinking, ‘Gosh, streaming works!’ That was the beginning of our great, global expansion.”

From then on, Hastings bet the farm on streaming. A year later digital revenues had reached $1.5 billion (£950 million) and the company had launched into Latin America and the Caribbean. The UK, Ireland and Scandinavia followed in 2012.

Lean and silver-haired, with a goatee and an easy-going, laconic manner, Hastings, 54, sits in Netflix’s European headquarters, which overlooks a canal in the heart of Amsterdam. The space is so new that it’s almost entirely empty, although the artwork has been finished: graffiti art adorns the walls, and doors have been decorated with unnerving, life-size blow-up portraits of characters from some of the channel’s best known series, including a malevolent George “Pornstache” Mendez (played by Pablo Schreiber) from Orange Is the New Black. Hastings is in the Dutch capital to host a dinner with journalists to mark Netflix’s first anniversary in the Netherlands, and to oversee the firm’s push into Europe.

France was one of six territories into which Netflix launched in autumn 2014 (the others were Germany, Austria, Belgium, Switzerland and Luxembourg), as growth in the core US market, where the service currently has 37.2 million customers, inevitably starts to slow. (Hastings has said that, in the long term, he’s aiming for between 60 million and 90 million subscribers in the US.) Next on his list are central, eastern and southern Europe. “We’ve got Asia, the Middle East and Africa still to go,” he says. “Our basic view is that we want to make Netflix available everywhere in the world.”

Netflix’s success has startled an industry whose rules — from infrastructure to distribution — have changed only slowly since the advent of digital. Legacy companies are even now still adapting.

“It’s time to remove all barriers to those who want HBO,” said Richard Plepler, CEO of HBO, one such competitor, when he announced to shareholders in October that the premium cable and satellite network, which has around 114 million subscribers worldwide, will launch a stand-alone online streaming service in the US in 2015.

Almost simultaneously, CBS — the venerable US network founded in 1927 — announced that it, too, was getting into the streaming business. CBS — which, in 2008, opted not to join ABC, Fox and NBC when they launched the ad-supported online streaming service Hulu — debuted CBS All Access, which charges $5.99 (£3.75) a month.

Other players, including Sony, DirecTV, Showtime, A&E and Starz, are likely to enter this increasingly congested marketplace, thereby joining a host of incumbents ranging from Amazon Instant, Hulu and France’s CanalPlay to sports services such as ESPN and Sky Go, and long-established competitors including the BBC’s pioneering iPlayer and Channel 4’s 4oD. Even the worlds’ biggest e-commerce site, Alibaba, is deal-making in Hollywood, as it, too, muscles in on VOD with its own set-top box.

“Netflix is the necessary figure in what will come to be seen as the year that television staged a jailbreak,” wrote New York Times columnist David Carr of 2014. “Yes, Netflix, which was supposed to lay waste to traditional media companies, may have saved them instead.”

According to Neil Hunt, Netflix’s chief product officer, the game-changing element behind the company’s success was its investment in streaming and picture quality. Hunt, a Briton with a softly transatlantic accent who is based at Netflix headquarters in Los Gatos, California, explains how rather than laying pipes itself, the company partners with “hundreds” of local ISPs — such as Verizon, AT&T and Comcast in the US; Bouygues Telecom and Orange in France; Deutsche Telekom in Germany; Virgin Media and BT in the UK — in order to deliver its content directly to customers. (Netflix’s arrangements with ISPs vary. With some, including Bouygues, Orange, Deutsche Telekom, Virgin Media and BT, it’s integrated with set-top boxes. Verizon, AT&T and Comcast, on the other hand, are simply carriers, which charge Netflix a toll for carrying their network service.)

Hunt explains that to ensure streaming quality, Netflix has purpose-built its own content-delivery network (CDN) called Open Connect — essentially, a set of servers that house Netflix video that the company offers to ISPs either to install directly into their own networks, or to connect with at common internet exchange points throughout the world.

“If the data is travelling a shorter distance — the industry calls it the last mile — it makes things work substantially better,” he says. “But in many ways, the real smarts happen on the client end. We leverage very straightforward, conventional protocols — in this case just http — to gather data very efficiently and use it as effectively as possible. A key piece of that is adaptive streaming. We can pick and choose which of several different versions of the video and audio to request and deliver, in order to get the best picture that your particular internet connection is capable of. We spend a lot of time trading off how quickly and how aggressively it shifts up to higher-quality levels, and how much buffer to retain, so it never stops, never freezes and doesn’t interrupt unless there’s absolutely no alternative.”

Netflix’s second differentiator lies in its use of data, terabytes of which are analysed to build recommendations for each subscriber. Every title on the service is tagged and cross-referenced with a vast array of global viewer data, including what an individual subscriber watched and how they watched it. Did they binge watch seven episodes in a row, for instance? Did they fast forward through certain sections, or rewind to rewatch a particular scene? Or if they abandoned a film, when did they do so? And what did they click on next?

According to Kevin Slavin, assistant professor and founder of Playful Systems at MIT Media Lab, this has far-reaching implications for legacy television’s business and delivery models. “The premise of lining up entertainment around a demographic, which is basically what networks do, may fall apart when you place it next to lining up entertainment around a person, which is what Netflix does,” he says. “It’s not just the shift in availability, or mode of consumption, but it’s also the shift in how decisions get made about what to watch. It’s true that one of the single biggest determinants about whether a show is going to get watched on linear network television is what was on right before it. But if you’re given an alternative [way to choose] that seems to be vastly preferable.”

Data is also used to guide Netflix on how to spend its programming budget, which was $3bn (£1.9bn) in 2014. Most of it is spent on licensing, but a growing, undisclosed percentage goes to commissioning original series. Award-winning shows such as House of Cards and prison drama Orange Is the New Black have been critical and ratings hits (Netflix doesn’t release viewing figures). It has also
co-produced two eight-episode seasons of the Norway-set mob comedy drama Lilyhammer, starring Steven Van Zandt (Silvio in The Sopranos), as well as the fourth season of the celebrated Fox show Arrested Development and a new ten-episode drama, Marco Polo, released in December 2014.

Last September, the company announced it had acquired the streaming rights to Batman spin-off Gotham from Warner Brothers and, two months later, commissioned The Crown, a series about the life of Queen Elizabeth II. Netflix is venturing into movie production too, backing and releasing (online and in selected cinemas) a sequel to Crouching Tiger, Hidden Dragon and four Adam Sandler films.

In all these cases (indeed, with every show it acquires or commissions), the company uses subscriber data for what it terms “audience prediction”. Multiple factors such as genres, cast and director, and how prequels have fared (if they exist) on other platforms/channels, are analysed to boost likely success. With House of Cards, for example, it knew from customer viewing habits that a political thriller starring Kevin Spacey and directed by David Fincher would appeal to a very large slice of the Netflix audience.

Content crosses borders, and international expansion is a key part of the Netflix strategy, says media analyst Michael Nathanson of MoffettNathanson. “In many countries, you’ve got lower pay-TV penetration than in the US and a streaming service like Netflix is a real alternative,” Nathanson says. “Also, because Netflix needs to programme more original content, having scale in contract acquisition globally helps them become more efficient.” The company’s potential customer base is about seven times larger outside the US.

But although expansion across Europe and beyond means rapidly rising subscriber numbers, new customers are expensive, Nathanson adds. He forecasts an international loss of $193 million (£122m) in 2014 and $166 million (£105m) in 2015 and estimates that it will be 2017 when Netflix’s international operation becomes profitable. “They’ve said very clearly that it’s going to cost them money to expand,” he says. Not everyone welcomes Netflix. Film producers in France claim that it avoids paying into the compulsory national film-subsidy fund by opting to make Amsterdam its base.

“Under EU law we’re a Dutch company and we’re governed by Dutch law, so we don’t pay into that fund,” Hastings says. “We’ve been open about that. It’s also true that some local French firms don’t like that. But that’s an EU integration question, right? [And, anyway] it doesn’t hold back the consumer. The main thing is just getting great content. The more content we get, the faster we grow… [But] it’s really internet video that’s growing. YouTube is growing by huge numbers, Hulu is growing. BBC iPlayer is growing. We’re propelled by this great tailwind of people having strong enough broadband, fast enough to be able to click and watch.”

Broadband speed isn’t the only tailwind Hastings has been able — and prescient enough — to take advantage of. His decision to switch to streaming has dovetailed with a number of other important industry-wide changes. Incumbent television companies were slow to see the potential of VOD, viewing it as a niche service that would be dwarfed by legacy broadcast delivery. Indeed, the same year Hastings launched a streaming-only service in Canada, Time Warner CEO Jeffrey L Bewkes was asked whether he viewed Netflix as a threat. “It’s a little bit like, is the Albanian army going to take over the world? I don’t think so,” Bewkes replied. Hastings’s nose-thumbing response was to buy a consignment of Albanian army hats and issue them to staff to wear around the office.

Growth has also been driven by what Netflix describes as “content connoisseurs”. Wanting to avoid expensive cable or satellite multichannel “bundles”, these tastemakers previously sourced high-quality content wherever they could find it, often illegally. According to a Netflix insider, these influencers came to the firm early and helped it grow, by telling their friends on social media what they were watching.

But it’s really in the area of data mining that Netflix has been able to leverage both first-mover advantage and technological edge. Personalised recommendations, for instance, save viewers from having to sift through hundreds of titles in the Netflix catalogue as they search for something to watch. Similarly, the tool means the service isn’t restricted to offering members its latest or most popular content. It uses data to find an individual viewer’s Netflix “doppelganger” — someone, who could be anywhere where the platform is available, who also feasts on weekend binges of The Walking Dead and maybe an episode or two of Orange Is the New Black, plus wildlife documentaries on Sundays. Netflix analyses the titles a doppelganger has watched that the first viewer hasn’t and offers those. This sort of personalisation is always iterating, Hunt says. From 2007 to 2009, Netflix held open competitions to crowdsource improvements to its predictive ratings algorithms, which resulted in the award of $1 million (£600,0000) to a team that devised an algorithm named “Pragmatic Chaos” (although it was never implemented).

“Since then we’ve come a great deal further,” Hunt says. “With a streaming service, we get a lot of signals about what and how people are watching and enjoying. For example, if you start something and abandon it in a few minutes, that’s a pretty negative signal. If you watch an episode actively to the end, with occasional pauses and rewinds, then power down your device, you were clearly there and engaged all the way through. Plus we know what we’ve shown to you — we know what we put on the screen as possibilities for you, what you snapped up or passed over in favour of something else. All of those of signals, put together, allow us to use a variety of algorithms to produce Top Picks lists, Popular on Netflix lists, and also to produce the micro genres, the rows of things like ‘Featuring a strong female lead based on a book’.”

Hunt prefers not to think that he and his team (which is “about 1,000 strong”) are building different channels for categories and micro-categories of users, but rather that they’re designing 50 million channels, one for every user.

“So it’s incredibly personalised, and we learn a lot about individual taste and customise something for each individual,” he says. “We learn a lot about the content and also by looking at things like the co-viewing frequency matrixes — which help us to understand which things are similar to which other things with certain groups of people; Markov chain models; matrix factorisation; and a variety of machine learning tools that help us to make ‘if this, then that’ type conclusions.”

The entertainment industry has barely scratched the surface of what’s possible when it comes to data. “The evidence for that is that pretty much every week we make discoveries about how to make things incrementally better. We did about 300 A/B tests, where we built two or three different versions of how to present a particular choice, or how to calculate what comes in a ‘post-play’ screen, over the past 12 months, and about 150 of those, about three a week, led to material improvements in the metrics we measure, so the number of hours people watch, or their attention, or other metrics we care about improving. By tiny increments, we make more progress. There are no areas we explore and can’t improve. Everywhere we look there’s room to make it better.”

Algorithms to drive recommendations and other improvements to service delivery are uncontroversial. But some commentators have wondered how they might affect the creative process. Hastings insists that the artists Netflix commissions have complete freedom. “We want the artist to have creative vision and not to assemble a bunch of sugar candy or car chases…” he says. Breton confirms this: “Normally you have the pressure of [an executive] saying ‘No, no, yes, no’. But Netflix says: ‘You are the producer, he’s the writer, you are the ones who know. So just do the job the best you can.'”

As the competition among streaming services intensifies, Netflix has been forced to firefight on a number of fronts. The company has been battling a number of American ISPs, including Verizon and Comcast, which are levying fees for Netflix — which now accounts for over a third of American internet traffic, at peak times — to interconnect with their networks and provide the higher-bandwidth connections necessary for better streaming quality.

“It’s not obvious who should charge who,” Hunt says. “You hear a lot of posturing about the ISPs spending a lot of money building up/out their infrastructure and that Netflix should pay its share. But the reality is that Netflix has paid its share — we’ve paid hundreds of millions of dollars to bring our content right to the front door, as far into the internet as we can, and the ISPs’ customers have paid for that last-mile bandwidth, so there’s clearly a difference of opinion there.”

Eventually Netflix struck deals with Verizon, Comcast, Time Warner Cable and AT&T to “pay the toll”, as Hastings puts it. But it is, he acknowledges, a long-running issue, which is unlikely to disappear. “It’s settled for a couple of years, but then it’ll come back, right?” he says, adding that he expects to face similar bills in Europe too. No wonder then that Netflix is lobbying to portray ISPs charging interconnect fees as damaging the very notion of a free and open internet, with equal access for all. “We’re super positive [about net neutrality],” he says. “We’ve lobbied for no fast lanes, no slow lanes. If you have enough speed then you don’t have to get into the fast lane or slow lane.”

A single-lane, superfast connection is, of course, integral to Hastings’ vision for the next frontier for television, which he believes centres on both internet streaming — which he predicts will account for around 50 per cent of all viewing in the US by 2020 — and on 4K Ultra HD television (with a resolution of 3,840 x 2,160 pixels). This is the first format that’s internet-centric, rather than broadcast-centric, he says. The shift from standard- to high-definition TV has been slow because it didn’t make sense for traditional broadcasters to change their signal when only five per cent of consumers owned an HD TV set, he says. “And there was no incentive for consumers to buy an HD TV if they couldn’t get any content on it — so that chicken-and-egg problem forestalls the development of better audio-visual standards.

“But the internet solves that, because we can now have one in a thousand people buy a 4K television, and then Netflix can supply them, over the internet, with 4K content. So then you can, much more quickly, move from one in a thousand, to one in a hundred, to one in ten, to almost everybody, because you don’t have to switch over the whole broadcast spectrum. You can treat each individual as an individual. So we’re very big on 4K Ultra HD television, which is the next big wave, and you’ll see TVs now out there at £1,500, and coming down. Samsung and Sony are both being very aggressive in pushing that — and we’re their primary source of 4K content.”

Maybe so. But in a competitive marketplace. Disrupters inevitably get disrupted. Is there a force Hastings fears? A rare — albeit brief — pause ensues. “We worry about movies and TV shows getting disrupted,” he replies. “As long as there’s a strong market for movies and TV, we should be fine – nothing’s going to out-internet the internet. So really, it’s just that one day people might start looking at movies and TV as kind of like the novel: boring 19th-century stuff. And that in 15 years, there’s Google Glass, combined with virtual reality and a morphine drip, and [when] you put all that together, it’s an amazing experience, so you go: ‘Watch Mad Men? Really?’ That’s the risk,” he says.

James Silver wrote about the £7,000 computer in 04.14

This article was originally published in WIRED magazine (UK) 02.15