Netflix Is Watching You Watch
And they aren’t shy about admitting it
Netflix sits in a league with the titans of today’s technology landscape. Like Google, Facebook, Amazon and the rest, its business model relies on endlessly analyzing users’ behaviors and preferences. The depth and sophistication of these operations is growing by the hour.
It’s old news that our consumption habits are constantly hoovered up by the biggest digital platforms, but few companies seem so publicly indiscreet about it as Netflix. Take their recent explainer on the implementation of a new approach to user-specific artwork:
The very short version is that Netflix has worked out how to get you to watch (and, hopefully, finish) a video by figuring out the artwork you’re most attracted to, and most likely to click. Are you a bigger fan of Uma Thurman than John Travolta? Based on your viewing habits, Netflix knows this. So when Pulp Fiction pops up in your feed, it’ll be decorated by a picture of Uma.
That doesn’t sound too nefarious, and certainly isn’t surprising. After all, data-driven engagement has always been backed up by complex collection techniques. Often these methods involve parsing patterns abstracted from the behavior of millions of users. Use one curatorial approach with one group of people, a different approach with another group, see which yields the most traffic and go with that one. Big data from big numbers.
But as Netflix explains in their article, they’re aiming for a level of unprecedented personalization, both in terms of how content is presented and how data is analyzed. Using visual analysis, they’re aiming to appeal to a market of one. That is, rather than simply judging the effect of one or another bit of artwork on a large group of people, Netflix is analyzing the components of individual images that resonate with individual viewers. It’s a visually-driven approach to data collection that leverages our aesthetic sensibilities to drive engagement.
In their article, Netflix reveals that they employ some untold number of designers whose job it is to craft artwork according to specific preferences. Data about how people respond to different visuals, actors, and other particulars are fed to the design team, who then craft artwork to be placed in front of specific users. Of course, doing this for each individual customer is a task far too large for human curation to keep up with. That’s why machine vision, they say, will play a key role.
Today, machines are capable of discerning and even generating images with a startling degree of subtlety, and of unpacking rich semantic information from them. So it makes sense that a service like Netflix — whose product is almost entirely visual — would explore ways of using that ability to maximize engagement. Visual search, facial recognition, augmented reality, all rely on the increasing ability of machines to see and make sense of what they see. And developments in these areas stand to make a huge impact on human/machine interactions; as Pinterest CEO Ben Silbermann recently told FastCo: “the camera is the next keyboard.”
It’s worth mentioning that these visually-driven technologies also present open channels for mining data from users themselves. Facial expressions, eye movement, a user’s environment, all stand to be observed as a gauge of how people respond to various signals. After all, the latest iPhone is not equipped to read your face in real time just so you can send animated emoji.
Many people are uncomfortable with the idea that their TV is listening in on their conversations, or their phone watching their face, whether with the intent of holding their attention or of selling whatever can be gleaned from the resulting data. Yet Netflix seems unconcerned with the optics of their own data collection operations.
Taste aside, the above tweet got an unusually high level of engagement. A large proportion of the responses were amused, even taking part in the gag.
However one feels about it, this is probably something we can expect to see more of. After all, it’s not the first time a data-driven content platform has fed its user data into marketing efforts. But not everybody finds the implications of this sort of thing entirely amusing.
Facebook certainly seems to understand that people are wary of being watched and manipulated by the platforms they use. But the domestic snooping carried out by the biggest tech companies is generally justified both as a means of keeping their services affordable, and of driving engagement (i.e. keeping customers hooked). This relates to the ‘dopamine feedback loop’ conversation revived by a Facebook exec this week, warning that social media is essentially tearing society apart.
In any case, it makes a certain kind of sense that the endless interplay of content curation and consumption would extend to social media. It’s a place where sentiment can be tested and shaped, and reflects the mechanisms of endless returns that Netflix and basically every other digital platform seek to establish with their users. We probably won’t know how effective any such efforts actually are, as Netflix is known to be quite tight-lipped about these numbers, but that’s another story.
Perhaps enough people have expressed sufficient ease at being subjects of constant analysis that their behavior can now serve as fodder for catty corporate social media editors? Maybe Netflix has the data to indicate as much? Could it be that Tweets like the one above are part of a broader effort to assess public sentiment over their being watched? Related to Trevor Timm’s first question, that might help explain why Netflix’s PR desk has access to such granular information about individual users’ viewing habits. It might also explain why Netflix is so openly sharing the gritty details about how they study our behavior and tastes…
As machines grow ever more adept at making sense of images, semantics, and personal sentiment, none of the above is out of the question. That this activity is now the subject of snarky Tweets suggests privacy concerns have done little to temper the trend. In the end, perhaps all of this could — and should — incentivize those who aren’t comfortable with being constantly studied in depth by their content providers to be still more vocal about it.
