film / tv / politics / social media / lists / web / celeb / pajiba love / misc / about / cbr
film / tv / politics / web / celeb


How Does Netflix's Algorithm Choose Recommended Content For Viewers?

By Jodi Smith | Streaming | April 2, 2018 |

By Jodi Smith | Streaming | April 2, 2018 |


I’ve been watching a lot of horror with my daughter the last few days. Then we watched all of the second season of Stranger Things on Saturday. I watched the Werner Herzog documentary Into the Abyss at least a month ago, if not longer. I’ve started and stopped Veronica, Temple, Darkness Rising, and The Woman with 7 Personalities over the last couple of months. So how does the Netflix algorithm use all of that information to tell me that because I watched Into the Abyss that I might enjoy watching the Walt Disney Short Films Collection?

Some of the other recommendations make sense, like Deadfall and Lost for Life, but Disney Short Films and a documentary about beekeeping seem a bit off. I understand that nothing is 100% foolproof, but what is the mysterious process that generates my Netflix home screen to look the way it does while my husband’s screen is so different?

According to Wired, the process uses metadata, thousands of “taste group” categories, explicit and implicit data, and the cooperation of human workers tagging shows for machines to analyze and categorize based on another set of information.

Did you know that there are Netflix workers that watch every single television show and movie offered by the streaming giant and that they tag the media based on its content? This leads to tags like “Ominous TV Shows,” “Violent Movies,” “Absurd TV Comedies,” and “Binge-worthy TV Thrillers” — all categories showing on my Netflix homepage. These make sense to me, as they obviously know I will destroy hours watching Mystery Science Theater 3000 or the entire season of The Confession Tapes.

So the tags are added to the shows and movies and then the computerized portion of the process helps Netflix decide how to weigh the tags, when specific tags were viewed, and whether something was binged or watched over the course of days. All of that is analyzed and processed for the more than 250 million viewer profiles that use Netflix. The data is then broken down into specific categories and populated with movies that may pique the viewer’s interest or you may get “Because you watched Chill with Bob Ross flanked by recommendations.

But wait, there’s more.

Netflix has you pegged and they know that I want to see mayhem, darkness, and the occasional ridiculous displays of absurdity. This information is used to present each recommendation in a way that will make you click on it. As Thrillist explains, your habits decide what scenes from the show or movie will appear in the thumbnail of the provided choices.

I watch a ridiculous amount of horror, documentaries about crimes, dark dramas, superhero movies, and ridiculous television shows. My husband watches stupid rom-coms, documentaries about the government, and other random stuff. My home screen shows for the title character angrily perched on her desk for Jessica Jones while his shows Jessica kissing some dude. The Gift thumbnail for him is Jason Bateman and Rebecca Hall while mine is Joel Edgerton looking creepy and holding a gift. It knows what we will likely click on and what we will skip based not only the description of the media but the image that they use to present it. It would be a little stalker-y if it wasn’t so damned helpful for discovering new shows and movies to watch.

Jodi Smith is a Senior Reporter, Film & Television at Pajiba. You can email her or follow her on Twitter.

Michael Rooker Lives the Dream of Every Q&A Moderator Ever | How Did Jared Die on 'The Walking Dead'?