Skip to content Skip to navigation

A New Day in Hollywood: Season Two, Episode Four of Raw Data

 
Attention is the currency of the internet: the ad-driven model of web publishing requires clicks and page views to drive money toward content and its creators. One problem with the model is surveillance; advertisers want to know as much as platforms will tell them about their users, to better target their ads. The information advertisers can glean about a site visitor ranges from browsing history to device information to personal details shared with, or on, the platform, as on Facebook. Another problem is the free rider problem; ad blocking software provides all the content with none of the capitalist intrusion, but deprives platforms of what is often their only source of revenue. 
 
Just as print media has struggled with the transition to digital, movies have struggled to transition to digital formats, trying multiple ways of restricting the playing and copying of media (and inducing viewer frustration) and experimenting with how to sell content online. The subscription model used by Netflix, Hulu, and other streaming and DVD-renting services has proven viable, and Netflix has avoided introducing third-party ads to its system, but Netflix is still subject to the problems of surveillance (tracking what you watch and when, how much you like it, and your binge watching patterns) and free loading (from shared accounts to Netflix-original shows ending up on pirate streaming sites). 
 
As noted in the episode, Netflix is fighting internet conglomerates like Amazon and Google to keep you within an ecosystem. Netflix promotes binge-watching by auto-playing episodes and continually tinkering with what media is available and how it is categorized; Amazon and Google try to use your behavior in some parts of their ecosystem to customize their offerings in other parts (e.g., YouTube videos recommended to you based on your Google searches, or Amazon videos recommended based on your purchases). Will Netflix branch out into other lines of business to gather more information about viewers? Will it begin offering a pay-per-view option the way that Amazon or Google Play do, for titles available on its platform through the DVD mailing subscription but not streaming? 
 
The path forward for a media platform involves several layers of the type of data-driven decision-making described in this episode. One layer involves what content to produce or procure the rights to; not only does a platform have to decide what has been popular, it also has to predict what will be popular, and what new movies and TV shows its users will want to see. Netflix’s advantage is that in a movie studio, decisions about how to license the back catalog aren’t made by the same people—or even the same department—as decisions on what new products to green-light, and Netflix can produce synergies to its advantage (such as producing a sequel or spin-off to debut when it knows it will acquire streaming rights to the original). A second layer involves how to pitch the content on the platform to the platform’s users. Knowing how to recommend media based on past viewing behavior is a perennially difficult problem; Amazon relies on an algorithm that believes you will like a second toaster more after having bought one toaster, and algorithms that attempt to sort movies into categories can paint themselves into dead ends (there are only so many martial arts-based interracial buddy cop movies) or recommend Homeland to someone who liked Legally Blonde (smart, troubled blonde protagonist in a job where others doubt her competency uses her unique insight and people skills to save the day?). 
 
The third layer of data-driven decision-making is in growth strategy, where an analysis of viewer behavior might recommend additional business lines (want a bottle of wine mailed with your DVD every week?) or suggest that a pay-per-view option would generate further revenue (how many streaming-only customer searches result in movies only available on DVD?) The dilemma for the model Netflix has chosen is that it’s not making more money off of those binge-watchers than it is off of someone who watches one TV show episode a month—or who doesn’t watch at all, but has forgotten to cancel his subscription. In fact, Netflix’s bandwidth costs to offer that high-resolution no-lag streaming will increase the more its subscribers become heavy users. Should Netflix then optimize for content that is just interesting enough to keep you coming back, but not so interesting that you stay for hours? Its decision to continually rotate what is available to stream each month points to an understanding that online attention follows a gambler’s model; users will keep checking back for the chance that something really great is available to watch, even if most nights they never actually find something they want to watch. (How many times have you spent more time looking through Netflix’s categories than actually watching something?)
 
These attention games are relatively harmless when they’re only about what entertainment we watch (and traditional television plays many of the same games) but what about attention diversion techniques that are designed to keep you from asking questions? Next week, Cyber Initiative researcher Jennifer Pan explains how social media in repressive regimes is designed to toe the party line. 
 
Listen on iTunes, Soundcloud, or at http://worldview.stanford.edu/raw-data