Latest / The Operator's Manual with Fexingo: Daily Decisions, Tactical Wins, and Business Execution / How Netflix Turned Data Into Content
Transcript
- Lucas: Three thousand two hundred. That's the number of television shows that were on the air in the United States in 2012. By 2025, that number had more than doubled. And in the middle of that explosion sat one company that bet its entire business model on the idea that data could tell you which shows would work before anyone watched them. Luna: You're talking about Netflix, and the story everyone loves to tell is how they used data to greenlight 'House of Cards'. But I think there's more to that story than the headline. Lucas: Much more. The 'House of Cards' decision is almost too perfect an anecdote — it's become a kind of creation myth for data-driven content. But the real lesson is how Netflix built a system that turned viewer behavior into a competitive advantage, and then had to rebuild that system when the game changed. Luna: Let's start with the famous part. What did Netflix actually know before they commissioned 'House of Cards'? Lucas: In 2011, Netflix was still primarily a dvd by mail company with a streaming side business. They had about 24 million subscribers in the US. But they had something more valuable than subscriber count: they had every single data point about what people watched, when they watched, what they paused, what they abandoned, what they rewatched. Luna: So they knew that users who rented the original British 'House of Cards' also tended to watch movies directed by David Fincher and films starring Kevin Spacey. That's the famous Venn diagram. Lucas: Exactly. The data showed a significant overlap between those three audiences. Netflix's chief content officer at the time, Ted Sarandos, has said that the model predicted that a show combining those elements would be a hit with a high degree of confidence. So they committed $100 million for two seasons — a huge bet for a company that had never produced its own content. Luna: And it worked. 'House of Cards' became a cultural phenomenon and legitimized Netflix as a studio. But here's the question I have: was the data actually the decisive factor, or did it just give management the confidence to take a risk they were already leaning toward? Lucas: That's the right question. And the answer is probably both. Look, every studio has hunches. Warner Bros. had a hunch about 'Friends'. HBO had a hunch about 'The Sopranos'. What data did was quantify the hunch and attach a probability to it. It allowed Netflix to make a bet that seemed irrational to traditional TV executives — two-season guaranteed order, no pilot, creative freedom — but felt rational internally because the numbers backed it up. Luna: So data became a decision-making tool. But Netflix didn't stop at greenlighting. They used data throughout the production and distribution process. Lucas: Right. The most visible example is the personalized artwork. Netflix realized that the same poster doesn't work for everyone. They started A/B testing thumbnail images — showing different users different artwork for the same show. If you see a poster with the lead actor's face, that's because the algorithm predicted you respond to that actor. If you see a dramatic scene, that's because data said you prefer emotional hooks. Luna: I've heard that they test up to a dozen different thumbnails per title. And that can increase viewership by double-digit percentages. Lucas: Yeah, it's a massive lever. But the deeper layer is how data influenced the creative decisions inside the show. For example, Netflix knew that a significant portion of their audience watched shows in binges — two or three episodes in a sitting. So they structured episodes with cliffhangers at the end of each episode, but also with a subtle 'reset' at the beginning of the next one, in case someone watched over several days. Luna: That's fascinating. They literally engineered the narrative pacing based on viewing behavior. Lucas: They did. And it worked spectacularly for a while. By 2020, Netflix had over 200 million subscribers globally. They were spending $17 billion a year on content. And the data engine was humming. But then something happened. Luna: The 2022 subscriber loss. They lost 200,000 subscribers in the first quarter, and the stock dropped 35% in a single day. Lucas: That was the moment the narrative shifted. Critics argued that the data-driven approach had led to a glut of content that looked similar — algorithm-friendly shows that were 'good enough' but rarely great. The phrase 'Netflix content' became almost pejorative, meaning something that was polished but formulaic. Luna: So the data that had been their superpower started working against them? Lucas: In a way. The algorithm is good at optimizing for the average. It can tell you what most people will watch. But it's not good at predicting what will become a phenomenon. 'Squid Game' is a perfect example — a Korean survival drama that was a massive hit globally. Could Netflix's data have predicted that? Probably not. It was a creative risk that paid off because someone had a vision, not because the numbers said so. Luna: And that's the tension at the heart of data-driven content. The algorithm can tell you what worked before, but it struggles with novelty. Lucas: Exactly. So Netflix had to pivot. They started giving showrunners more creative freedom. They reduced the number of original series and focused on quality over quantity. They also started sharing more data with creators — something they had previously been notoriously secretive about. The thinking was: if we give our partners the same insights, they can make better creative decisions. Luna: I remember reading that they started providing detailed audience retention data to producers — like exactly where viewers drop off in an episode. Lucas: Right. If the first 10 minutes of episode 3 have a 20% drop-off, the showrunner knows something isn't working. That's incredibly valuable feedback. But it also creates pressure to conform to what the data says works, which can kill originality. Luna: So how do you balance that? How do you use data without becoming a slave to it? Lucas: Netflix's current approach is what you might call 'guided intuition.' They use data to identify opportunities and reduce risk, but they leave the creative decisions to humans. For example, when they decided to make 'The Crown', data showed that period dramas had a loyal but relatively small audience. But they believed the show had the potential to break out. So they went ahead, and it became one of their most acclaimed series. Luna: So data as a compass, not a map. Lucas: That's a good way to put it. And it's a lesson that extends far beyond streaming. Any business that uses customer data to make product decisions faces the same challenge. The data tells you what your existing customers have done. It doesn't tell you what new customers want, or what your existing customers might want if they saw something different. Luna: Let's talk about the specific mechanics. How does Netflix actually collect and process all this data? I mean, billions of hours of viewing every year. Lucas: They have a data architecture that captures every interaction — every play, pause, search, scroll, rating. That data flows into a recommendation engine that uses collaborative filtering, content-based filtering, and deep learning models. The system generates personalized rankings for each user based on their history and the behavior of similar users. Luna: And they have to do this for over 200 million accounts, each with their own profile. The computational scale is enormous. Lucas: It's one of the largest machine learning systems in the world. But here's the thing: the algorithm isn't just about recommending content. It's also about deciding what content to acquire or produce. Netflix uses a 'test kitchen' model for some shows — they'll commission a small number of episodes, see how they perform, and then decide whether to order more. That's a direct application of the lean startup methodology to content creation. Luna: And that's where the data-driven approach really shines — in iterative learning. You can test a concept, get real user feedback, and adapt. Lucas: But there's a catch. The feedback loop takes time. And in the streaming wars, time is expensive. Disney+, HBO Max, Apple TV+ — they all launched with massive libraries and big-budget originals. Netflix couldn't afford to iterate slowly. They had to place big bets quickly. Luna: So data helped them place those bets, but it also led to an arms race in content spending. By 2025, Netflix was spending over $20 billion a year on content. Lucas: And that brings us to the current state. The streaming market is maturing. Subscriber growth is slowing in developed markets. Netflix is now focusing on profitability and free cash flow rather than just subscriber adds. And data is playing a different role now — it's about retention, not just acquisition. Luna: What do you mean by that? Lucas: Netflix's data scientists have built churn prediction models that identify subscribers at risk of canceling. They can then target those users with personalized recommendations, content reminders, or even special offers to keep them engaged. It's a retention play that's entirely data-driven. Luna: So the same data that helped them choose shows now helps them keep people watching those shows. It's a full circle. Lucas: Exactly. And the key takeaway for any business is this: data is not a substitute for judgment. It's a tool that amplifies judgment. Netflix succeeded because they had leaders who understood both the power and the limits of data. They were willing to make bold bets that the data supported, but also to override the data when creative instinct demanded it. Luna: So what's the next frontier for Netflix's data strategy? Where do they go from here? Lucas: I think the next big shift is incorporating more real-time data. Right now, Netflix's recommendations update periodically — typically every 24 hours. As they move into live programming — like the Chris Rock special or sports events — they'll need to adapt recommendations in real time based on what's happening. That's a completely different technical challenge. Luna: And there's also the question of privacy. As regulations tighten and third-party cookies disappear, Netflix's first-party data becomes even more valuable — but also more scrutinized. Lucas: That's a huge point. Netflix has one of the richest first-party data sets in the world. They know what you watch, when you watch, where you pause. That's incredibly powerful for advertising — which they're now entering with their ad-supported tier. But it also means they have to be extremely careful about how they use that data, both legally and ethically. Luna: So the company that once said 'we compete with sleep' now has to compete with data privacy regulators too. Lucas: Yeah. And that's the ongoing challenge of building a data-driven business. The tools get more powerful, but the constraints get tighter. The winners will be the ones who can navigate both. Luna: Any final thought on what other businesses can learn from Netflix's journey? Lucas: I think the biggest lesson is to start with a clear strategic question, not with the data. Netflix didn't start by saying 'Let's collect all the data and see what happens.' They started by saying 'We need to decide which shows to produce, and data can help us make better decisions.' The data served the strategy. Too many companies do it the other way around. Luna: Data as a servant, not a master. That might be the most important insight of all. Lucas: And it applies whether you're producing a billion-dollar TV show or deciding which product feature to build next. The tools change, but the principle stays the same.