On Netflix’s linear programming experiment #

‘Direct’, the company’s new linear channel, available to French subscribers via the web player, isn’t Netflix’s first attempt to solve its lean-back experience problem. Rather, it’s a continued commitment to integrate the one thing traditional TV did right – creating a no-effort experience.

This same commitment is what brought us autoplay, which was a revolution at the time and various ‘shuffle mode’ experiments. With autoplay, Netflix made it easier to continue watching than to stop, by removing barriers to a viewer’s psychological inertia. The existence of the term Netflix binge is testament to their success. once you’ve chosen what to watch, you’re hooked. If a viewer can’t choose what to watch though, autoplay is powerless.

Shuffle and linear programming are Netflix’s efforts to defeat decision paralysis. Presenting someone with many options frequently overwhelms their decision making ability, resulting in indecision. This effect is compounded when there is little difference between the outcomes of a particular choice, or when a person is unable to know and remember all available choices. I'd hate to know just how many hours are lost trying to choose what to watch.

To my mind, both Netflix’s experiments miss the mark for two reasons:

  1. Shuffle only works for episodic shows. Dropping randomly into a serial is like starting a movie from the middle. Confusing.
  2. Linear programming should still be tailored to me. I don't have the same interests as the rest of the population, even when I don't know what I want.

A better solution might be something akin to radio stations. Instead of picking from infinite shows, cut that decision down to picking from a select set of ’moods’; irreverent comedy, crime drama, action anime etc. Once a mood is picked, a viewer can play and skip shows as they wish. Skipping picks another show that fits the mood, otherwise autoplay plays the next episode of the series. Every show always picks up from where you left off, we shuffle shows not episodes. Finally sprinkle in machine learning as needed, for personalisation within each mood.