3 ms·

Having listened to podcasts with several important Netflix people as guests, I have picked up that they are unintentionally user hostile. They're so focused on

• vl 7y ago
  The problem is that optimizing for long-term satisfaction metrics is extremely tricky and hard. Such metrics lead to the better user retention, potentially even with less watch time. Paradoxically higher watch time may lead to lower user satisfaction (and decreased retention). In ML world these metrics require combining DNNs with methods like Q-learning. YouTube learned these lessons and now probably has the most advanced ML among competition in its’ recommendation system.
  • goalieca 7y ago
    Watch time is bullshit. Focus on making a quality product that people enjoy. My goal in life is not to spend any more time in front of the tv than 1-2 hours per day. Usually it’s just before bed when I relax.
• closeparen 7y ago
  Stated and revealed preferences are very different. In some ways this is the fundamental problem of media. It killed newspapers, too: once they could analyze readership at a per-article level, it became clear that people who say they value true, in-depth reporting are liars. No one reads that stuff. Content increasingly became shaped by what the data say people actually want, despite it being fashionable to talk about one's disgust for it. I imagine Netflix is thinking along these lines. Ratings are performative; engagement metrics tell the truth.