Support for LAist comes from
Audience-funded nonprofit news
Stay Connected
Audience-funded nonprofit news
Listen
Podcasts Take Two
How does Netflix's algorithm know what you want to watch?
solid orange rectangular banner
()
Aug 9, 2013
Listen 5:49
How does Netflix's algorithm know what you want to watch?
Have you ever taken a look at your Netflix's "Suggested for You" queue? It's an algorithm-generated list that's supposed to take all the stuff you've watched and tell you what else you might like.
SAN FRANCISCO, CA - OCTOBER 24:  Netflix envelopes sit in a bin with other mail at the U.S. Post Office sort facility on October 24, 2011 in San Francisco, California.  Online movie rental company Netflix reported third quarter earnings of $62.5 million, or $1.16, per share compared to $38 million, or 70 cents per share one year ago.  (Photo by Justin Sullivan/Getty Images)
Netflix already knows what you're watching, but how does it know what you should be watching?
(
Justin Sullivan/Getty Images
)

Have you ever taken a look at your Netflix's "Suggested for You" queue? It's an algorithm-generated list that's supposed to take all the stuff you've watched and tell you what else you might like.

If you're like most people, you probably use Netflix for watching shows and films you already know you like or ones in tandem with your interests. Depending on how often you use it, though, there's a good chance that Netflix already knows what you like, too. 

Don't believe us? Try watching the first selection in the "Suggested for You" queue.

It's an algorithm-generated list that's supposed to take all the stuff you've watched and tell you what else you might like. We won't tell you what our list looks like, but if you're a Netflix user, you've probably wondered how that algorithm works.

Wired writer Tom Vanderbilt wondered, too, so he tracked down the guys who track your entertainment preferences.

The process consists of two steps: first is the team of human taggers who watch movies and categorize them. Second is the algorithm that takes those categories and fits them to your viewing preferences.

How can you make sure that you're getting recommendations that you enjoy? It's simple. According to Vanderbilt, all you need to do is watch more movies.

"Like any of these systems, the more information you give it, the better it works," he says, "I like to call it 'training your algorithm.'" 

Over the years, Netflix has evolved its algorithm to make training it easier for the customer. While you still have the option to rate titles after you watch them, Vanderbilt says that it isn't all that important anymore. Apparently, there's quite a disparity between what people rate and what they actually like.

"It turns out, the best way to predict what someone will want to watch is what they're actually watching, not what they're rating," said Vanderbilt. 

So, if your recommendations are based off your viewing habits, why are some of them rated two or three stars? There's a reason for that, too.

"[Netflix] wants to throw in a certain amount of variability," said Vanderbilt. "A lot of us have these guilty pleasures hidden somewhere in the queue. Your 'Hot Tub Time Machine' and stuff like that."

So, even if you're watching the same genre every night, Netflix is there trying to make sure you don't get bored. And even if you do get tired of watching your recommended titles, there are still other rows like 'What's New' or 'What Are Other People Watching' so you can discover new titles.