Verified by 
Key points
- Netflix's commitment to scientific testing helps the company refine its offerings.
- The company deploys a number of measures that reflect an understanding of cognitive biases.
- Netflix's user interface is loaded with personalized elements, driving user engagement.
- Advancements in technology allow the possibility for more deeply personalized content.
Netflix continues to grow from strength to strength, commanding 270 million subscribers who watched 183 billion hours of content last year.
A Culture of Testing
Netflix executives “…believe there’s a better way to make decisions about how to improve the experience we deliver to our members: we use A/B tests…Instead of small groups of executives or experts contributing to a decision, experimentation gives all our members the opportunity to vote, with their actions...”
This gives an insight into Netflix’s awareness of the perils of groupthink and a raft of other biases that can beset boardroom decision making. Netflix notes the weight of charisma in such decision-making processes, rather than scientific evidential backing, evoking Weber’s concept of charismatic authority. True to a behavioral mindset, this approach places empirical evidence over cognitive assessment, or charisma-infused persuasion.
Netflix is amongst a host of other successful organizations carrying out thousands of experiments to refine their offerings, including Google and Amazon. Organizations have recognized the need to evaluate with scientific experimentation; this conclusion is a founding principle of behavioral science. Latent within these corporations is a distinct consumer focus. This might reflect some of the parallels between the fields behavioral science (which emerged from the work of those such as Daniel Kahneman), and behavioral psychology (which emerged at the beginning of the 20th century, through the work of those such as John Watson and B.F. Skinner). Psychologist and historian of psychology Dr. Brian Hughes roots the emergence of behaviorism in the socio-political upheavals of the time, as well as the emerging demand to persuade masses of people, and shape their consumptive behaviors. Organizations like Netflix have been able to excel in a way that the old behavioralists could not because of the scale and volume at which they can conduct scientific testing, enabled by advances in technology.
Netflix regularly employ a number of testing routes, including A/B testing. More intriguingly, they also rely on Contextual Banditry; this form of experimentation relies on using data that is known to make a live decision about which intervention will best work for which customer. Netflix has also detailed its effective use of quasi-experimentation when other routes are not available or are unsuitable.
Behavioral Tools
Beyond this commitment to scientific testing, Netflix has integrated knowledge of a raft of cognitive biases to their offering:
1. Salient Reassurance. Netflix has obviously identified a key barrier to signing up for their free trial: concerns about committing to a contract that cannot be broken. They directly address this by hammering home the point that subscription can be cancelled at anytime. While signing up, I counted no less than four times that the fact I could “cancel anytime” was highlighted. A system 2 approach to this would inform that the customer only needs to be told once to understand this basic point. An understanding of system 1 instead informs that repeating this critical message is crucial.
2. Allusion > Full Disclosure. Netflix trialed allowing potential customers to see their whole catalogue in the hope that this would encourage more signups. This had an unintended effect: Signups did not increase. Catching this unintended effect captures precisely the value of scientific evaluation. Netflix then decided to give customers a peak into the catalogue, and pepper that with allusions to volume, without disclosing the full amount. The visual carousels they employ are one way to achieve this.
3. Disclosure > Unsettlingly Accurate Yet Opaque AI. Interestingly, Netflix realized that hard, mechanistic personalization might unsettle customers, so they decided to begin disclosing some of the thinking behind personalization, to good effect. This disclosure reassures customers, but also further develops their brand image as being a relatable and amicable friend, rather than a faceless and calculating corporation. Disclosures are made in the form of “because you watched…” and “others who liked…” statements.
4. Top-10 Effect. In 2020 Netflix deployed a “Top 10” list on its homepage; this might be due to the “Top 10 Effect” which proposes that people mentally divide masses of information into more digestible categories. More recently, it began publishing top-10 lists across countries and genres. The weekly updated list would also evoke salience of certain titles: Knowing that others are watching a certain title would also increase the likelihood of us watching it, by way of social norms and social proof.
5. Visual Engagement and “Moments of Truth." Netflix is (in)famously known for their use of auto-playing videos; merely hovering over a title elicits commencement of its trailer. This is purposeful, and is designed to keep the viewer visually engaged; Netflix believes it has a window of 60-90 seconds to get the viewer committed to watching something or risk losing them to a competitor. They term this window the “moment of truth”; the focus on continuous stimulation is to avoid “idleness aversion”.
6. Default Bingeing. “Idleness aversion” is related to another of Netflix’s famous practices, and among the most well-known successful applications of defaulting: the auto-playing of videos that encourages bingeing. It's a simple, inexpensive default that has impacted on the behavior of millions.
Personalized User Interface (UI)
We might have grown up being told to “never judge a book by its cover”, but behavioral research into the framing effect (and indeed the messenger effect) increasingly shows that many of us do just that. Netflix has combined its commitment to testing with its understanding of framing to deploy different thumbnails to exhibit the content they offer.
How do they decide which image to show which customer? Using the contextual bandit approach, they consider “the titles they’ve played, the genre of the titles, interactions of the member with the specific title, their country, their language preferences, the device that the member is using, the time of day and the day of week”
Recommendations are also personalized by Netflix; the strength of this system is such that as many as 80% of shows watched are through its recommendation system, rather than users going through a deliberative search of content.
Netflix personalizes homepage displays, but also search algorithms and messaging and marketing content. As well as artwork and trailers, Netflix even personalizes Dynamic Sizzles—the “montage of video clips from different titles strung together into a seamless a/v asset” that users are presented with. For each user, the Dynamic Sizzle’s contents are informed by that user’s top titles. The higher ranked a suggested title is for that user, the more it is weighted when creating that user’s personalized Sizzle.
When strict personalization is not possible, Netflix has identified and segmented 2,000 ‘taste communities,' or segments of their viewers who had similar viewing preferences. The process of segmentation allows Netflix to individualize what is offered to their viewers in groups. While it falls short of personalization, it remains more effective than a traditional ‘one size fits all’ model.
When personalization and segmentation are not appropriate, Netflix can fall back on broader approaches such as ‘nationalization’, wherein certain geographic regions receive different outputs. As an example, users navigating within the UK will see certain titles featured in the background, while those in the Middle East will see others, usually local content.
Personalized Content
Going beyond merely realizing intricately personalized UIs, could personalized content be the future?
Netflix’s boldest venture towards content personalization came by way of their 2018 title Bandersnatch which went as far as allowing each viewer to decide the path forward taken by characters in the special; could this be a glimpse towards the future of entertainment media consumption? True to Netflix’s experimentational form, some have suggested this was a masterstroke in decisional data mining.
As revolutionary as Bandersnatch was in allowing individual users to make decisions about the way in which the story progressed, the story arcs were restricted to a set number. The speed of advancement within the field of AI could soon open up new horizons in this regard. Turning again to their content, in 2023’s most recent season of Black Mirror, we see another glimpse into what the future might hold as what begins first as fiction paves the way for an incoming reality. In the episode “Joan Is Awful” a fictional character’s life is suddenly portrayed within a series on Black Mirror’s in-series dystopian version of Netflix, “Streamberry”. The level of insight into this individual’s life was incredibly intrusive, while the speed at which the production was realized and made available to Streamberry’s viewers was unnerving.
While the idea of fully personalized content might seem unlikely, consider how personalized much of our lives are growing to become, as the age of the self emerged. Consuming media content began as a social exercise shared with fellow viewers in the cinema at restricted times, towards becoming a more solitary exercise in choosing to view specific content on demand in our own dwellings. Social media has personalized media in a way that would not have been immediately predictable a century ago, as we all create and upload content that reflects our lives and our interests, and consume content that is tailored to our personal inclinations, often bolstered by personalizing algorithms.
Conclusion
Netflix’s affordance of primacy towards a culture of testing is indicative of a commitment to science, while its deployment of behaviorally informed facets to its product reflects a growing mastery of the way the mind works. Advancements in technology have allowed the company to make unprecedented moves in the realm of personalization, and the exciting and unnerving possibilities of this guiding content is emerging. The company’s approach is colored by psychology; this has been one of the reasons for its growing success and represents a model for the use of science to create a high-powered product that is continually refined in such a way as to engage its users.
This post is an update on a previously published post on the London School of Economics’ Psychological and Behavioural Science Blog
