Not many readers are likely to be familiar with Captain Power and the Soldiers of the Future. Prior to reading sociologist T.L. Taylor’s 2018 book Watch Me Play: Twitch and the Rise of Game Live Streaming, I certainly wasn’t. Having grown up in Yorkshire in the 1990s, my opportunities to travel back in time to watch a 1987 children’s TV show broadcast only in Canada
and the US were limited.
Captain Power is worth discovering, however.While the show’s plot trod familiar sci-fi archetypes(“Earth, 2147. The legacy of the Metal Wars, where man fought machines – and machines won”), its format was unusual. “[M]erging content and equipment, [Captain Power] offered a special toy for engagement,” writes Taylor. “Children who had purchased the ‘Powerjet XT-7 Phoenix’ [light gun] were able to ‘fire’ at the TV and carry out live battles.” Captain Power wove the typical merchandising that surrounds children’s television into the act of viewing itself – you bought the toys and the toys fed the content. It won’t surprise you to learn that Captain Power was produced by Mattel.
The Powerjet was designed to look like a spaceship and equipped with sensors which detected glowing markers that appeared on screen during battles in the show. Children shot at these markers and, in turn, the villains “fired back”. Get hit by enough of these blasts1 and the Powerjet’s auto-eject was triggered, launching a tragic figurine of Captain Power from its cockpit. “Are you going to help Captain Power and the Soldiers of the Future™?” blared a poster campaign for the toy. “Or are you just going to stand there?”
Captain Power was cancelled by its paymaster Mattel after a single season, in part because of the unrelenting crapness of the Powerjet XT-7 Phoenix.2 Nevertheless, after reading Taylor’s assessment of the programme, I became fixated on the idea behind its interactivity. A television show where input from the viewer shapes what you see on screen, I thought,
settling down to a night of Netflix and chill.3 _I wonder if anyone’s done anything with that since?
In March 2019, Netflix UK began doing strange things with its pricing. While a standard subscription to the streaming platform was supposed to cost £7.99 per month, new visitors to the site started receiving a range of quotes that seemed entirely dependent upon the web browser they were using. A handful of Chrome users were quoted £8.99, while those on Internet Explorer were told the service would cost £9.99.4 Meanwhile, Netflix’s £9.99 premium subscription showed up as anywhere between £10.99 and £12.99. The really weird thing, however, was that whenever anyone actually did subscribe to Netflix, those rates dropped back to the standard fee at checkout. “We are testing slightly different prices to better understand how members value Netflix,” a spokesperson for the company eventually explained. “Not everyone will see this test and we may never roll out these specific prices beyond this test.”
Two months later, Netflix announced sweeping price hikes in the UK. It wasn’t a surprise. The company had already introduced the biggest price increase in its 21-year history in January 2019, with subscription costs climbing between 13 to 18 per cent across the US, Caribbean and Latin America. The UK increase ultimately fell within a similar range – standard subscriptions rose to £8.99 per month, while the premium plan went up to £11.99. “We change pricing from time to time as we continue investing in great entertainment and improving the overall Netflix experience,” explained a company spokesperson of the US increase. While Netflix was going to cost more, a UK spokesperson reassured customers, it would remain “great value for money compared to other options on offer”.
Except, that’s probably not true.
In 2017, one of Netflix’s major content partners, Disney, revealed that it was pulling its programming from Netflix by 2019. Disney’s films and television shows had been featured prominently on the platform since the two companies struck a deal in 2012, a development that Netflix’s chief content officer Ted Sarandos labelled “a game changer”. Disney, however, had now decided to go it alone, with CEO Bob Iger announcing plans to build “a direct-to-consumer Disney service”. This platform, Disney+, was revealed in April 2019 and is scheduled to launch in the US on 12 November 2019. Disney+ will cost $6.99 a month, and become the online home for all of the studio’s films and television shows. Stocked with current productions from franchises and studios such as Star Wars, Marvel and Pixar, the platform will supplement its existing content with freshly commissioned television series, as well as an ongoing effort to digitise the entire Disney back-catalogue.5 “We are all-in,” announced Iger at the service’s launch, in which he emphasised the company’s “treasure trove of long-lasting, valuable content”, which “no other content or technology company can rival”.
Not that that will stop them trying. Around the same time as the Disney+ announcement, Apple revealed its own plans to enter the subscription streaming market with Apple TV+, a service which will feature original content from Steven Spielberg, J.J. Abrams and Oprah Winfrey among others. While Apple argued that “this is not just another streaming service”, it absolutely is – and a sizeable one at that, with an estimated $2bn budget for creating original programming. “The TV+ product plays in a market where there’s a huge move from the cable bundle to over-the-top [streaming media offered as a standalone product],” said Apple’s CEO Tim Cook. “We think that most users are going to get multiple over-the-top products and we’re going to do our best to convince them that the Apple TV+ product should be one of them.” Apple has not announced costs for its platform as of yet, although rumours suggest it will be priced at the same $9.99 a month as the company’s existing Apple Music service.6 If this proves true, Apple TV+ will be $3 per month cheaper than the equivalent Netflix package.
Between new rivals Apple and Disney (as well as existing streaming services such as Hulu, Showtime, Sling TV, YouTube TV, Amazon Prime Video and HBO Go), Netflix suddenly finds itself in heated competition for dominance of an industry that is expected to be worth $124.57bn by 2025. “Very few entities in this world can afford to spend $200 million on a movie,” said Alan Horn, chairman of Walt Disney Studios, in Anita Elberse’s 2013 book Blockbusters: Hit-Making, Risk-Taking, and the Big Business of Entertainment. “That is our competitive advantage.”
Erm, has he seen how much money Apple has?
Whenever I click on Netflix, I head to the “Top Picks for Oliver” section,7 an area of the website in which the platform curates programming based on your past viewing habits. It’s always interesting to see what the service makes of you. When I logged on today, my eye was drawn to a new Top Pick – a small image of a disapproving woman in a large formal hat, emblazoned with the logo Sherlock Gnomes. Intrigued, I hovered over the image, which expanded into animated footage of garden gnomes investigating a crime scene. I couldn’t help but look closer.
“After their friends disappear from gardens all over London, Gnomeo and Juliet team up with super sleuth Sherlock Gnomes to help find them.” I liked this synopsis, which stretched across the screen the moment I clicked affectionately on the gnomes. An additional pop-up box further reassured me, providing information about the film’s cast, genre, and core description (“This film is: Exciting”). So, I started clicking through Sherlock Gnomes’s 1h26m running time. There were several sections I skipped; a handful I went back to for a second look;8 and some in which I hit pause so I could google additional information about the movie (“Sherlock Gnomes is sadly, utterly stumped by the mystery of the reason for its own existence,” Rotten Tomatoes). When I’d had enough, I closed the window, safe in the knowledge that I could return later if need be. I’ll always have “Resume”.
Sherlock Gnomes was added to Netflix UK in February 2019, with the film’s streaming rights having been purchased from Paramount Pictures as part of Netflix’s estimated $15bn annual spend on developing and acquiring new content. This budget is allocated either towards the creation of original productions, or else funnelled towards external studios to acquire streaming rights to their pre-existing content – all of which is paid for by Netflix’s subscription fees and external investment. It’s a business model termed “bundling”, which is impractical for most physical goods given manufacturing costs, but which thrives when applied to digital products. “Digitisation makes it easier and more profitable for producers to sell entertainment goods in large bundles than what would be possible with physical goods, and bundling information in this way creates significant economies of scale,” write Michael D. Smith and Rahul Telang in their book Streaming, Sharing, Stealing: Big Data and the Future of Entertainment. “The more products the seller has in a bundle (say, the 10,000 or so shows on Netflix), the better the seller can predict the average value of the bundle across consumers. Not everyone assigns the same values to the same movies, but in a large bundle the differences in the individual values average out.” For the record, it will take a hell of a lot of excellent content to balance out Sherlock Gnomes.
Bundling seems to be working for Netflix – at least up to a point. While the company is reported to be anywhere between $8bn and $20bn in debt, depending on which news source you listen to, it’s still the clear industry leader in terms of both exposure to the general public and ongoing annual investment. Netflix has more than 65 million subscribers worldwide (as well as additional viewers accessing the platform through friends, family and illegal proxies), and accounts for around one-third of North American internet traffic. These numbers are high, albeit commensurate with broader changes in the way society consumes content. A 2013 Harris Interactive study found that “nearly 80 percent of US adults with internet access watch TV through subscription on-demand services” like Netflix, and this percentage has continued to grow. Meanwhile, the average age of traditional television audiences is rising as younger viewers abandon the medium to source content from elsewhere.9 “One in four millennials is a ‘cord cutter’ [those who have cancelled subscriptions to cable television],” note Smith and Telang. “One in eight is a “cord never’ [those who have never paid for cable television].” As to where younger viewers are going instead, Smith and Telang are unequivocal: “Online.”
Except, maybe we shouldn’t be calling them “viewers” any more. In his 2002 essay ‘Watching the Internet’, theorist and artist Dan Harries coined the portmanteau “viewser” to describe a new category of consumer who experiences content as a “hybrid mode of both viewing and using”. While Harries wasn’t specifically discussing subscription streaming, “viewsing” is as good a term as any to capture what happens on platforms such as Netflix. Within a streaming service, the subscriber is encouraged to think of themselves as active. “Watch TV on your schedule,” promises one of Netflix’s marketing campaigns; “Spend quality time together,” reads another, “with no interruptions.” Exploiting entrenched ideas of traditional viewers as passive receptacles for content,10 Netflix bills itself as returning agency to the audience. A subscriber is not simply a viewer, but also a user – able to click through the site to select the programme they want to watch, and then further interact with the platform as they determine the manner in which they want to watch it. There is no set schedule; no commercials; and, ever since the platform launched its post-play system in 2012 (in which credits are cut short to automatically commence the next episode of a series), no delays in programming. As the advert proudly boasts, “no interruptions”.
Unless you fancy one. Bathroom break, anyone?
At the 2016 Consumer Electronics Show (CES) in Las Vegas, Netflix CEO and co-founder Reed Hastings took to the stage in front of a packed auditorium. It was a triumphant moment for someone who had spent years being dismissed as a parvenu by traditional television networks and film studios. “It’s a little bit like, is the Albanian army going to take over the world?” quipped Time Warner’s CEO Jeff Bewkes when asked about the threat posed by Netflix to existing media networks in 2010. “I don’t think so.” Speaking to The New Yorker writer Ken Auletta in 2014, Hastings revealed that he had been aware of Bewkes’s remarks. “For the next year, I wore Albanian Army dog tags around my neck,” he said. “It was my rosary beads of motivation.” On stage at CES, Hastings began pacing back and forth. “Entertainment and technology are continuing to transform each other, as they have been doing for over 100 years,” he told his audience. Behind him, a video screen started glowing with renders of national flags from across the world. “Today, right now, you are witnessing the birth of a global TV network,” Hastings proclaimed, announcing that Netflix had simultaneously launched in 130 new countries, taking its global total up to 190.11 In an act of considerable restraint, Hastings did not display the Albanian flag. “Whether you are in Sydney or St Petersburg, Singapore or Seoul, Santiago or Saskatoon, you now can be part of the internet TV revolution,” he said. “No more waiting. No more watching on a schedule that’s not your own. No more frustration. Just Netflix.”
At least in its public pronouncements, Netflix paints itself as a company that is consciously redesigning television. In particular, it wants to take an axe to television “flow”, an idea developed by the cultural critic Raymond Williams in his 1974 book Television: Technology and Cultural Form. “In all developed broadcasting systems the characteristic organisation, and therefore the characteristic experience, is one of sequence or flow,” wrote Williams. “It is evident that what is now called ‘an evening’s viewing’ is in some ways planned by providers and then by viewers as a whole; that it is in any event planned in discernible sequences which in this sense override particular program units.” Unpacked a little, Williams’s point is that traditional television is not consumed as discrete programming, but rather as a succession of different segments – including advertisements and news breaks – with this stream in its entirety amounting to the actual object of viewing. “Watching television,” observes Djoymi Baker in her essay ‘Terms of Excess: Binge-Viewing as Epic-Viewing in the Netflix Era’, “actually meant watching television flow, not watching a particular program.”
The prevailing design principle behind Netflix and its ilk is to break this idea of flow, or else replace it with a new form of sequencing through the application of digital-media techniques. In some senses, this change is modest: the removal of adverts and a break to the hegemony of television scheduling does not represent a sea change to the nature of video in the way in which the user-generated content of YouTube does. Netflix’s programming remains traditional narrative-driven television and film, all of which is studio-produced or backed. “We are not a generic ‘video’ company that streams all types of video such as news, user-generated, sports, porn, music video, gaming, and reality,” reads the company’s ‘Long-Term View’ document. “We are a movie and TV series entertainment network.” This sense of focus, and the element of televisual tradition embedded in it, clouds the relationship between streaming and network broadcasting. “Other than being delivered via IP, Netflix had almost nothing to do with the conventions of digital media – in a sense it rejected them,” argues essayist Michael Wolff in his 2017 book Television Is the New Television: The Unexpected Triumph of Old Media In the Digital Age. “It is not user generated, it is not social, it is not bite size, it is not free. It is in every way, except for its route into people’s homes[…] the same as television.”
Audiences, however, don’t agree. “The distinction of the Netflix user experience is such that some younger viewers perceive the service as other than television, even if they watch Netflix on a television,” notes Emil Steiner in his essay ‘Binge-Watching in Practice: The Rituals, Motives, and Feelings of Streaming Video Viewers’. “Viewers’ ability to watch on multiple devices [and] their technological control of content[…] were identified as essential to the experience.” It’s a sensation of control that Netflix’s interface has been specifically designed to engender. “Until 2015, the Netflix desktop interface had a lightgrey background,” observes Ramon Lobato in his 2019 book Netflix Nations: The Geography of Digital Distribution. “Video artwork was formatted in vertical, DVD-style boxes, so that the overall effect was reminiscent of a video store.” Under the platform’s current design, however, the screen has been darkened, with imagery of its content now running horizontally – as if “suggesting frames on a celluloid filmstrip,” writes Lobato, who also detects a visual reference to the dark of movie theatres in the platform’s new black background. Certainly, Netflix’s decision to stud its image grid with autoplay trailers of headline films and series (inevitably those produced by Netflix itself) captures something of the theatre of cinema. “There’s a droid here,” hisses a tearful Hilary Swank, her A-list face beamed across my screen to promote the launch of Netflix’s new sci-fi film I Am Mother. Right on cue, an expensively CGId robot fills the screen, standing in the doorway with a kind of insouciant, Who? Me?
Beneath these trailers, the platform’s celluloid reels invite you to cycle through content quickly and easily, with the site’s superabundance of free-flowing imagery suggesting a surfeit of programming for the intrepid viewser. Each reel is labelled with an oddly specific category (“Irreverent US TV Programmes”; “Binge-worthy Criminal Investigation TV Programmes”; “Way Out There”; “Raunchy TV Programmes”), which seem to have been designed to suggest an intimate knowledge of the viewser and their tastes that runs far beyond the capability of any traditional retail platform. It is, Sarah Arnold writes in her essay ‘Netflix and the Myth of Choice/Participation/Autonomy’, a “userfriendly interface that maintains the perception of choice but also directs the viewer toward content more likely to keep them engaged and subscribed”. Lobato, meanwhile, sees the design as an effort to move “Netflix away from video-store and DVD culture – surely a fading memory for most of its users”, as well as to minimise its connections to television. “Interestingly, the iconography of television is nowhere to be found in Netflix’s interface design, despite the abundance of TV series available through Netflix,” he says. “There are no remote controls, advertisements, or schedules. Even though the idea of television is central to Netflix’s commercial ambitions[…] the television experience does not seem to be central to how Netflix wishes its users to imagine streaming.”
This same conceptual distance affects showrunners operating within the platform. In 2011, Netflix announced that it had secured a deal to license and distribute new episodes of Arrested Development, a cult sitcom that ran on Fox between 2003 and 2006. While the Fox iteration of the show followed a relatively traditional sitcom structure and focused on protagonist Michael Bluth’s efforts to reunite his dysfunctional family, the Netflix relaunch shattered this unified narrative perspective. Rather than proceed in a linear fashion, each instalment of the 15-part Netflix series covered the same period of time and the same events, with the episodes creating differentiation by taking up the perspective of individual members of the cast. Grouped together, the 15 segments run contiguously but narrate circularly, gradually filling in gaps in the wider story. “Not only will the episodes be available at the same time on Netflix, but they also cover the same period of time in the characters’ lives,” show creator Mitch Hurwitz told pop culture website Vulture. “I pretty quickly realised everything here is about the order of telling the stories, that there will be shows where you find out a little bit of information and then later shows where you revisit the scene and you find out more information.”
According to Cindy Holland, Netflix’s vice president of original content, this structural device was specifically developed for streaming and would have been difficult within the constraints of traditional viewing. “Part of the conversation early on is thinking about it as a 13-hour movie,” explained Holland in Clare Joyce’s 2013 essay ‘The cord-cutters’. Whereas traditional sitcoms are structured around weekly television schedules, rigid runtimes, and regular advertising breaks, Netflix is free from such considerations. Instead, the platform bills itself as being “about the freedom of on-demand and the fun of binge viewing”, consciously targeting the 70 per cent of US viewers who regularly watch multiple episodes in a sitting. To enable this, Netflix has adopted a number of design tricks. It began launching television seasons in their entirety as opposed to sequentially – a now-common digital strategy that originated with Netflix’s House of Cards in 2013 – as well as encouraging writers and show runners to work with the platform and its constraints in mind. “We don’t need recaps,” summarised Holland. “We don’t need cliffhangers at the end. You can write differently knowing that in all likelihood the next episode is going to be viewed right away.” “I would say that in its purest form, a new medium requires a new format,” said Hurwitz of his decision to take Arrested Development to the platform. “You can’t do in a short story what you could do for a novel, in a novel. You can’t do in haiku what you would do in a long-form poem. In a perfect world, we would be making something that could be only on Netflix, just as in years prior, you could make something that could only be on HBO.”
As research for this essay, I reached out to Netflix to ask whether anybody from the company might be able to speak to me about the platform and its design. “Hi Oli, Many thanks for getting in touch,” came a reply from the platform’s communications director. “I’m afraid that I don’t have anyone to put forward but this week’s interview with Ted Sarandos on the Media Show may be of interest as it covers some of these points. Thanks”.12 I listened to this interview and it was useless, so I tried again. Is there anybody on the team who might be able to speak to me about the design of the platform itself? It would be great to put Netflix in a design context and very helpful to have someone who could discuss this authoritatively. A return email came a week later. “I’m sorry, but it’s not something I have anyone available for right now. Thanks”.13
Just to be clear, Netflix saying it doesn’t have anyone available to discuss the design of its platform is like McDonalds saying it doesn’t have anybody available to discuss beef burgers or Shell not having anybody available to discuss oil – it’s not so much a lack of information, as a conscious decision to not share that information. I considered what might lie behind this reticence and came to a firm conclusion: Netflix has realised its staff sound terrifying whenever they discuss how the platform actually operates. Consider this quote from Netflix’s engineering director Xavier Amatriain, given to Wired magazine in 2013: “We know what you played, searched for or rated, as well as the time, data, and device. We even track user interactions such as browsing or scrolling behaviour.” Presumably keen to distil Amatriain’s message into even more sinister form, Sarandos has helpfully paraphrased this idea: “We have the viewing data of everything.”
Every action you take on Netflix is logged by the platform. The entire website has been designed as a data trap, such that any move you make triggers it. “Netflix doesn’t know merely what we’re watching, but when, where and with what kind of device we’re watching,” writes Andrew Leonard in his 2013 essay ‘How Netflix is Turning Viewers into Puppets’. “It keeps a record of every time we pause the action – or rewind, or fast-forward – and how many of us abandon a show entirely after watching for a few minutes.” Netflix knows that I watched Sherlock Gnomes for absolutely ages during working hours, for instance, and I fully expect an anonymous tip-off to Disegno’s publisher any day now. To quote Smith and Telang, “No movie or TV studio [has] ever been able to tap into so much detailed information about its individual consumers.” Netflix lives up to its promise of letting you watch what you want, how you want – in return, it just wants to know what that is, and how you do it.
Unlike Google and Facebook, the data Netflix generates isn’t monetised through advertising. Rather, it is fed back into the platform in two chief directions: user orientation and content creation. In his 2016 CES talk, Hastings specified that the platform’s ultimate aim is to “show you exactly the right film or TV show for your mood when you turn on Netflix.” Enabling this vision of beatific technological succour requires those two data directions to work in tandem. If Netflix can’t direct you towards the perfect programme within its back-catalogue, then it wants to create a new show that can scratch that itch. The platform is a digital colonoscopy, busy probing your deepest data to see what it can winkle out.
In 2013, Amazon’s vice president for digital music and video Bill Carr spoke to The Wall Street Journal about how exactly data shapes streaming. “We let the data drive what to put in front of customers,” he explained. “We don’t have tastemakers deciding what our customers should read, listen to, and watch.” It is this datafication of television and cinema that authors Kevin McDonald and Daniel Smith-Rowsey tackle in their book The Netflix Effect. “If there is a singular Netflix effect,” they write, “it may simply be that technology and entertainment are merging at an accelerating rate and seriously impacting the business and economics of mass media.” Smith-Rowsey takes the point further in his essay ‘Imaginative Indices and Deceptive Domains’, arguing that it is Netflix’s willingness to yolk digital data to creative content that has seen it emerge as “the definitive media company of the twenty-first century, perhaps best exemplifying the synergy and tension between Silicon Valley and Hollywood”.
In his 1983 Hollywood exposé Adventures in the Screen Trade, the screenwriter William Goldman gave short shrift to Hollywood’s commissioning process. “Not one person in the entire motion picture field knows for a certainty what’s going to work,” he wrote. “Every time it’s a guess and, if you’re lucky, an educated one.” His conclusion was clear: “Nobody knows anything.”
Goldman’s observation still rings true today. Television networks test programme ideas by commissioning pilots: one-off episodes that introduce a season and provide an opportunity for executives to gather audience feedback. It’s a costly enterprise. Producing a pilot for a drama series can run to anywhere between $5m and $6m, with some industry estimates suggesting that $800m is spent annually on failed pilots alone – those which never lead to a series. To decide whether a pilot is worth making, networks rely on audience sampling provided by focus groups, as well as ratings data and aggregate statistics purchased from third-party consumer research companies.14“In practical terms, this [meant that] the creative industries relied on ‘gut feel’,” explain Smith and Telang. “They could put together focus groups[…] but these were exceedingly rough measures based on tiny samples that were of questionable value when applied to the broader population. For the most part, the companies therefore had to rely on[…] people hired, optimistically, for their superior ‘instincts’.” As a result of this, traditional media outlets have limited direct interactions with their audiences. “We don’t have [a] direct interface with the American public,” admitted Sony’s CEO Michael Lynton in an interview given in 2014. “We need to go through an intermediary to do that.”
It’s a system to which old-guard executives still cling. “Data can only tell you what people have liked before, not what they don’t know they are going to like in the future,” argued John Landgraf, the president of FX Networks, in a New York Times interview from 2013. “A good high-end programmer’s job is to find the white spaces in our collective psyche that aren’t filled by an existing television show.” It’s a charming idea – human specialists holding out against faceless algorithms – but hopelessly outmoded.15 In 2011, Mordecai Wiczyk and Asif Satchu of Media Right Capital began pitching their new show House of Cards around television networks. The pair had based their programme on the original BBC drama of the same name, and already had commitments from Kevin Spacey to star and David Fincher to direct. The show was shopped around HBO, Showtime and AMC, where it met with a lukewarm response – networks were wary of investing in a pilot for a political drama, a genre traditionally seen as not playing well with viewers. The response from Netflix – which Wiczyk and Satchu had met with to discuss streaming rights following an initial television run – was different. Sarandos offered $100m up front for a two-season run of 26 episodes. No pilot was necessary. The show would go on to air for seven series, picking up seven Emmys and two Golden Globes, and is widely attributed with heralding a boom era for streaming. Immediately prior to the show’s launch in 2013, Netflix reported quarterly sales of less than $1bn. For the first quarter of 2019, that figure stood at $4.52bn.
“In many ways[…] the ‘risk’ of spending $100 million to create House of Cards was not a risk at all,” note Smith and Telang. The company’s offer was based entirely upon analysis of the viewing data it had gleaned from its platform. Sarandos had statistics to show that there was an audience that liked the original BBC show; an audience that liked the films of David Fincher; and an audience that responded positively to Kevin Spacey.16 “Netflix argued that it didn’t have to go through the standard pilot process, because it already knew from its data that there was an audience for House of Cards – and that it had a way of targeting potential members of that audience as individuals,” explain Smith and Telang. The platform developed multiple trailers, each of which highlighted a different aspect of the show and could be targeted towards the viewing habits of specific subscribers: those who liked Spacey got the Spacey trailer; those who liked “movies with strong female leads” got a version that went heavy on actors Robin Wright and Kate Mara; those who liked Fincher got footage showcasing the programme’s cinematic qualities. “As such, there was no gamble; the algorithms demonstrated that an audience existed for this program,” claim Smith and Telang. This has been the Netflix model ever since. “We don’t use data to influence creative at all,” Sarandos told the National Association of Television Program Executives in 2015. “Our data is mostly used to say ‘Wow, there’s a real there there for this show.All the elements are there for this to be a great big show, and therefore you invest heavily.”
Early on in Netflix’s exploration of streaming, the company’s founder Hastings suggested that subscribers might respond well to a “digital shopping assistant” – an in-platform helper who would have “a personality and a photo and could point customers to movies they would like in Netflix’s library”. Readers who have used Netflix at any point in the past 12 years will notice that this helpful assistant is nowhere to be found. Instead, as media scholar Neta Alexander observes in ‘Catered to Your Future Self: Netflix’s “Predictive Personalization” and the Mathematics of Taste’, Netflix discovered that “a nameless algorithmic system turned out to be a much more lucrative solution.” So much for personality.
Details of the current Netflix algorithm are kept vague, but it largely builds upon the company’s original recommendation algorithm, CineMatch. The basics of the system, at least, are clear. The content that a subscriber is presented with is determined by two chief factors:17 direct customer input gleaned from Netflix’s recommendation software, such as the taste preferences and show ratings that subscribers are invited to input; and user behaviour data drawn from observation of what people actually click on and watch. Within this system, not all data is equal. “The company distinguishes between user behaviour and user expression (of taste, interests and identity),” explains Arnold in ‘Netflix and the Myth of Choice/ Participation/Autonomy’. “It sees user expressions (via taste preferences and ratings) as poor data, as it doesn’t correlate as neatly with actual interactions and behaviour. The context offered by the user, namely the knowledge they produce about their personhood through wish lists and personalisation, is secondary to the knowledge produced by algorithms.” This isn’t something that the company shies away from talking about either. “[Most] of our personalisation right now is based on what [users] actually watch, and not what they say they like,” said Todd Yellin, the company’s vice-president of product innovation in a 2015 interview with The Verge. “[You] can give five stars to An Inconvenient Truth because it’s changing the world, but you might watch Paul Blart: Mall Cop 2 three times in a few years… so what you actually want and what [you] say you want are very different.”
To Netflix, this is an essentially benevolent system, with datafication removing commissioning power from the subjective tastes of television executives,18 and instead rendering it responsive to the viewing habits of its audience. For the first time, television audiences are getting exactly what they want, rather than what they’re told they want – or so the argument runs. There is also a shift in the intended subject of television viewing. Whereas traditional television targeted audiences – amorphous groups made up of bands of characteristics, attributes and identities – streaming promises to target the individual. When a platform has more than 10,000 shows on offer, it no longer needs to commission content that will hold a broad appeal for as large an audience as possible, but can instead target individual viewers with niche productions that respond specifically to their tastes. “Traditional television audience measurement was[…] somewhat speculative, ‘desperately seeking the audience’ but unable to locate or identify those outside of sample groups,” writes Arnold. “These new forms of measurement use data gleaned from online user interactions as a way of profiling and controlling the behaviour of every individual.”
Hastings argues that such changes fundamentally alter the viewing experience. “Think of it as entertainment that’s more like books,” he notes, his analogy tying in with broader cultural efforts to liken streaming platforms to the breadth of choice offered by a library, as opposed to the linear functionality of network television. “You get to control and watch, and you get to do all the chapters of a book at the same time, because you have all the episodes.” It’s a flattering comparison, and one with which media studies researcher Sudeep Sharma has rightly taken issue. “The service functions more like a newsstand [than a library, and] Netflix plays the role of ‘surrogate consumer’ for exhibitors,” he notes in ‘Netflix and the Documentary Boom’. “Rather than just simply providing access to texts, they are engaged [in] an effort to ‘push certain texts on consumers, rather than letting us pull what we want’.”
That’s not how Netflix would like you to speak about it, though. “Netflix posits the use of data mining systems as beneficial for the consumer and suggests that such systems allow the company to better understand and respond to audience tastes through its recommendation system,” writes Arnold. “This represents a shift in audience measurement and interpretation from the notion of the depersonalised mass to the personalised, the individuated, and the autonomous.” Speaking in 2009, Yellin had a folksier way of making the same point: “We are rolling out several features to delight our members with a more personalised website.”
Were any of you a bit weirded out earlier when I mentioned that Netflix’s House of Cards advertising campaign targeted viewers who liked programmes with “strong female leads”? Because, how could Netflix know it was the “strong female lead” that made you watch a show? And, how does Netflix know that you actually liked that show, even if you watched it all the way through to the end? People watch things they hate all the time. I finished Sherlock Gnomes, but that doesn’t mean I liked it. It doesn’t necessarily mean I like “strong gnome leads”.19
Given that Netflix’s entire philosophy is based around data, the company tends to give data-based answers to questions like these. While data gleaned from a single viewing experience might not tell the algorithm much, Netflix argues, the general pattern of a subscriber’s viewing data over time should begin to paint an accurate picture. “The more you use Netflix, the more relevant your suggested content will be,” explains the company’s online Help Centre. The problem with this kind of explanation, however, is that it doesn’t actually answer the question. In Raw Data is an Oxymoron, Lisa Gitelman and Virginia Jackson set out the fact that all data has an in-built interpretative element: “Data need to be imagined as data to exist and function as such, and the imagination of data entails an interpretative base.” Even if you watched 10,000 movies with a “strong female lead”, there remains an act of interpretation that has gone on to establish that those 10,000 films are movies with a “strong female lead”. Without that interpretation, there’s no data.
In Netflix’s case, these acts of interpretation are conducted by researchers working under the aegis of Yellin. “Using large teams of people specially trained to watch movies, Netflix deconstructed Hollywood,” writes journalist Alexis C. Madrigal in his 2014 The Atlantic essay ‘How Netflix Reverse-Engineered Hollywood’. “They paid people to watch films and tag them with all kinds of metadata. This process is so sophisticated and precise that taggers receive a 36-page training document that teaches them how to rate movies on their sexually suggestive content, goriness, romance levels, and even narrative elements like plot conclusiveness. They capture dozens of different movies’ attributes. They even rate the moral status of characters.” In other words, all of Netflix’s supposedly neutral data is based upon the judgment calls of teams of people trying to act in accordance with a 36-page corporate handbook.
Hoping to cast a little light on this, Smith-Rowsey spends a portion of ‘Imaginative Indices and Deceptive Domains’ mapping the categories by which Netflix orders its data. Searching through the platform, he lists 19 umbrella categories (things like ‘Action & Adventure’, ‘Drama’ and ‘Comedy’), around 400 subcategories (‘Military Action & Adventure’, ‘Social Issue Dramas’ and ‘Slapstick Comedies’), and about 73,000 micro-genres, which is where Netflix gets creepily specific. “Visually Striking Father-Son Movies, Violent Nightmare-Vacation Movies, Understated Independent Workplace Movies, and Emotional Drug Documentaries,” he writes. “Once, [cinema theorist] Rick Altman asked [film critic] Leonard Maltin to settle the question: is Thelma and Louise ‘a chick-flick, a buddy film, a road movie or something else?’ For Netflix, it is none of these, because Netflix has none of those three categories[…] [The platform] in effect privileges some films and shows and types of viewership, and to some degree re-constitutes what Netflix’s sixty million users think when they think of film and TV.”
Once you know this, Netflix’s professions to be directly shaping its platform in response to individual viewsers start to become unstuck. “[The] promise of personalisation and autonomy is undone since Netflix simply shifts such demographic markers to genre tags,” observes Arnold. “Identity is displaced from the user to the content. In other words, the user does not bring their identity – along with the complexities that inform it – to the platform, rather the platform has determined what these mean.” Because the one thing Netflix definitely isn’t doing is responding to you. At best, it’s responding to your data trail, or what the academic John Cheney-Lippold has termed an “algorithmic identity” generated by our behaviour online.
Cheney-Lippold’s algorithmic identity need have no real relevance to the person whose actions construct it. Its nature is simply a sum total of all behaviours within a set Netflix account (regardless of any real understanding of who has been using that account or, indeed, how many people have been using that account) as defined by the parameters set in place by Netflix’s metadata. “The knowledge produced via user digital interactions has no referent in the personhood of the user; their tastes, social values, or other non-digital behaviours and expressions,” notes Arnold. “[The] data mined by Netflix is not used to infer anything about the human agent interacting with the service; instead it finds correlations between profiles and data interactions.” Neta Alexander’s essay ‘Catered to Your Future Self’ finds an even more tragic way of summing up the impoverished interpretation of personalisation offered by Netflix: “[Subscribers] confuse the “You” in “Recommended for You” with a unique, complex individual rather than with a group of strangers who all happened to have made similar choices.”
Exactly 30 years on from the demise of Captain Power and the Soldiers of the Future, Netflix finally closed the circle. Launched on the platform in December 2018, Bandersnatch is a choose-your-own-adventure television show from the creators of Black Mirror, Charlie Brooker and Annabel Jones. Written by Brooker, the programme is a collection of some 250 segments that loosely tell the story of programmer Stefan Butler’s efforts to design a video game, and his subsequent mental breakdown. The exact nature of the story, however, and what the viewser actually sees on screen, is entirely down to choices made during the show itself. As you watch, you’re asked to make decisions that shape Butler’s life. These range from the small (should Butler eat Sugar Puffs or Frosties; should he listen to the Thompson Twins or Now 2?) through to the life-changing. Should Butler throw himself to his death from the balcony of a council block, or should that fate fall to his mentor, the video-game designer and acid enthusiast Colin Ritman?
On my first run through, I killed Ritman. I regretted it immediately. For someone so strung out on acid, Ritman spoke a lot of sense. “[It’s] a fucking nightmare world and the worst thing is it’s real and we live in it,” he told Butler at one point. “It’s all code.” Throughout Bandersnatch, that maxim is worth remembering. Prosaic choices that do little to shape its narrative are, in part, included within the programme’s runtime because they serve as training devices – picking a breakfast cereal amounts to an unimportant dry run to make sure the viewser is comfortable with the mechanics for when they eventually have to make a more significant decision. These trivial choices also add to the overall sense of control and freedom. You can even pick Butler’s breakfast! It’s a televisual sleight of hand, a trick designed to distract from the fact that every inch of Bandersnatch has been painstakingly mapped out and fitted into a rigorous, pre-planned structure. “There were points where in working stuff out, it got like trying to do a Rubik’s Cube in your head,” says Brooker of the writing process. “I literally had to get up from my desk and kind of walk around the house holding my head.”
Few viewsers will ever see all of Bandersnatch’s segments (“There are some things that are really hard to find in it,” says Brooker. “I couldn’t tell you how to get to them”), but a common theme across the myriad routes through the show’s plot line is Butler’s growing awareness of being controlled. What he believed to be his own free actions are, in fact, pre-determined eventualities steering him down pathways shaped from without. “I’m being controlled by someone from the future,” he cries at one particularly desperate moment in proceedings, when the viewser begins communicating with him through his computer monitor. Thankfully, you’re presented with the option to be straight with him.
You’re being controlled by someone on Netflix.