Violent Thrillers About Cats for Ages 8 to 10

January 2, 2014 9:52 AM

If you use Netflix, you've probably wondered about the specific genres that it suggests to you. Some of them just seem so specific that it's absurd. Emotional Fight-the-System Documentaries? Period Pieces About Royalty Based on Real Life? Foreign Satanic Stories from the 1980s? ... Through a combination of elbow grease and spam-level repetition, we discovered that Netflix possesses not several hundred genres, or even several thousand, but 76,897 unique ways to describe types of movies.
posted by Horace Rumpole (63 comments total)

This is a wonderful article, even if they missed a golden opportunity with the photo blurbed "Todd Yellin at Netflix headquarters." The part at the end about Perry Mason was wonderful; when I realized how intricate Netflix's genre system is, I knew something weird like that would happen and I am glad to know it did.

I wish there was an easier way to access these things on the user end, though. Like mix-'n'-match keywords for a genre and then get the movies in the genre. Once in a while I will watch, for instance, Highlander 3: The Final Dimension and it will magically unlock a whole category of godawful FX action films from the 90s that I have no way to access otherwise. On the one hand, it feels like Achievement Unlocked each time it happens. On the other, having direct access would mean I wouldn't be burned out after Highlander 3: The Final Dimension finishes and the genre magically disappears from the menu a day later.

Also the new profiles thing is great. Now my wife and I have separate profiles and both of our lives are easier now that my Gritty Action Movies of the 1980s don't get mixed up with her Violent True Crime Television Series.
posted by griphus at 10:05 AM on January 2, 2014 [8 favorites]

Because it is necessary to have this list on hand when discussing grand schemes of classification:
Celestial Emporium of Benevolent Knowledge

The list divides all animals into one of 14 categories:
Those that belong to the emperor
Embalmed ones
Those that are trained
Suckling pigs
Mermaids (or Sirens)
Fabulous ones
Stray dogs
Those that are included in this classification
Those that tremble as if they were mad
Innumerable ones
Those drawn with a very fine camel hair brush
Et cetera
Those that have just broken the flower vase
Those that, at a distance, resemble flies
posted by wotsac at 10:05 AM on January 2, 2014 [13 favorites]

A problem with the streaming service is that it seems to ignore any ratings applied to anything that isn't available on streaming, so most of the genre hits I get are less than helpful. I do love seeing all the mix-and-match tag descriptions and how they change every time I scroll through the list, and it's fun to read how the apparently-powered-by-Perry-Mason magic happens.

I'm terribly disappointed that I'll no longer get recommendations based on my interest in Breakin' 2: Electric Boogaloo. I can't even double-check to find out what genre tags it has anymore.
posted by asperity at 10:06 AM on January 2, 2014 [2 favorites]

The fake movies were my favorite part, but then I got to the part about Raymond Burr.
posted by box at 10:06 AM on January 2, 2014

Violent Thrillers About Cats for Ages 8 to 10

I've seen that one!
posted by capricorn at 10:15 AM on January 2, 2014 [5 favorites]

This is some fabulous (and fascinating) data wanking. Now I want to know what Netflix has for my preferences and interests, dammit, beyond the basic categorization.
posted by rmd1023 at 10:27 AM on January 2, 2014 [2 favorites]

"We wanted to highlight our personalization because we pride ourselves on putting the right title in front of the right person at the right time."

Ha ha ha no. Okay, first of all, according to what Netflix presents me with on my Roku, they have about 50 movies in total, about 10% are something I would ever or have ever watched and the rest of them are the same obscure action/crime garbage in the faded beat-up boxes at the mom-and-pop rental store that I didn't want to see in 1992.

Secondly, the theory of only presenting me with movies that have elements of movies I've already watched is deeply flawed. Quantifying which macguffins are being spoken of and how many of them were in movies you already saw is not a good way to figure out which movies you will enjoy in the future.

Thirdly, since it seems like they're not doing the star rating thing anymore, if you watch a bunch of movies you don't like, you're stuck in an infinite loop of misery because they're only going to show you what you've seen before and nothing else.

This seems like a lot of work to build a system that doesn't the work the way it should, and the way it should work is not what they set out to do. Fail.
posted by bleep at 10:27 AM on January 2, 2014 [13 favorites]

My bet is that in the early days of loading data, before they shook out most of the bugs in their tools, one of them got stuck in a loop and loaded tens of thousands of copies of just one Perry Mason show. I would love to be able to run a few strategic queries against their core database tables to see if this could be verified.
posted by benito.strauss at 10:27 AM on January 2, 2014

This is fantastic. Holy crap.

Raunchy Absurd Slashers

Heh.
posted by zarq at 10:28 AM on January 2, 2014

It's kind of hilarious to see this tagging project lionized as some kind of huge technical achievement rather than a boondoggle, given that the actual adjectives, as Netflix applies them to actual movies, are so completely ridiculous and vague and unhelpful. Once you get past the obvious and empirical, like "From the 1970s," it's pretty much all just unjustified critical handwaving on the part of the anonymous taggers — you see the same "quirky," "dark," "cerebral," "conspiracy" descriptors thrown at the 9/11 Truther Zeitgeist bullshit that you do applied to Noam Chomsky talking about international politics and Hitchcockian thrillers. It's as useless as I can imagine any such black-box recommendation scheme being — when I browse it mostly serves as unintentional amusement, like "Can you imagine they think these things make sense together?"
posted by RogerB at 10:29 AM on January 2, 2014 [1 favorite]

But I also realized there was a way to scrape all this data. I'd been playing with an expensive piece of software called UBot Studio that lets you easily write scripts for automating things on the web.

What? I mean, even I could probably write such a script, given a bit of time. He should have been pointed to Python, which is free.
posted by JHarris at 10:31 AM on January 2, 2014 [5 favorites]

I always had one type of movie or tv show I watched: movies with hunky male leads that I can watch on mute...every other category is unnecessary...
posted by Alexandra Kitty at 10:32 AM on January 2, 2014 [2 favorites]

How could they do that without making a way to click through to the 76,897 genres?

Playing around with the URL they give is fun, though.

http://movies.netflix.com/WiAltGenre?agid=666 "Dark British Independent Movies"

http://movies.netflix.com/WiAltGenre?agid=23333 "Visually-striking Girl Power Children & Family Movies"

http://movies.netflix.com/WiAltGenre?agid=11114 "Cult Comedies from the 1970s"

http://movies.netflix.com/WiAltGenre?agid=1974 "Dark Scandinavian Movies"

http://movies.netflix.com/WiAltGenre?agid=10011 "Critically-acclaimed Emotional Biographical Movies"
posted by fings at 10:33 AM on January 2, 2014 [3 favorites]

Yeah, I'm with bleep. I just went and checked; I've rated almost FIFTEEN HUNDRED movies since When you combine that with the hundreds and hundreds of movies I've rented since I got my account in January of 2005, I would expect that by now Netflix could throw me either movies that I've seen but haven't rated, or stuff I might actually enjoy. Instead it continues to give me the same bargain-basement shit that I used to ignore back when I rented movies at a brick-and-mortar store. This article was interesting, and I'd love for them to figure out a way to use the data they have, but I'm not seeing much improvement in their recommendations over the past decade.
posted by nushustu at 10:33 AM on January 2, 2014 [9 favorites]

I have a feeling that tonight's dreams will fit snugly into the "Nightmares about Todd Yellin's Poorly-lit Photograph" genre.
posted by not_on_display at 10:39 AM on January 2, 2014 [9 favorites]

There's gotta be a company out there that has some expertise in surfacing things you would be interested in based on your keywords plus an algorithm that weights crowdsourced and user-contributed metadata. If there is, they should help Netflix.
posted by stupidsexyFlanders at 10:39 AM on January 2, 2014

Quantifying which macguffins are being spoken of and how many of them were in movies you already saw is not a good way to figure out which movies you will enjoy in the future.

This is a good point, but one of the funniest things about the Netflix tag scheme as it stands — one of the things that makes it seem the most like film theory as reinvented by Engineer's Syndrome — is that it's not just about content at all, but a crazy hodgepodge of descriptors of form and/or content as if they were the same thing. It tries to recommend films that it thinks are like what you've just watched in effect and/or style as well as subject.
posted by RogerB at 10:40 AM on January 2, 2014 [2 favorites]

Playing around with the URL they give is fun, though.

My street address is "Scary Movies starring Christopher Lee."
posted by Shepherd at 10:46 AM on January 2, 2014

It tries to recommend films that it thinks are like what you've just watched in effect and/or style as well as subject.

I understand what it tries to do. But those things aren't what makes a good movie, or a movie I would be interested in watching. And it doesn't even accomplish it's stated task. I'm seeing this as a big problem from a user experience point of view, not as a harmless quirk or something as inexplicably delightful as a crappy 80's buddy cop romp. I also noted with interest that the team that built it was described as engineers, not as user experience professionals, information architects or content strategists.
posted by bleep at 10:47 AM on January 2, 2014 [1 favorite]

fings: How could they do that without making a way to click through to the 76,897 genres?

For a while, I've wondered if there was some addon or extension that would provide some way to go to the next page in a sequentially numbered URL, as these genre pages are listed. Here's the Firefox "Autopager" addon, which might do that, but I'm on a work computer, so I'll have to try it at home.

Oddly, you think Netflix might try to capitalize on these altgenres more themselves. After all, they have pages to browse DVD movies by genre and a similar page for streaming content (I think - the site is blocked at work, but I can view a 2012 version of the site through Archive.org), but those genre lists are less specific. For instance, under Sci-Fi & Fantasy you get:

• Action Sci-Fi & Fantasy
• Alien Sci-Fi
• Classic Sci-Fi & Fantasy
• Fantasy
• Foreign Sci-Fi & Fantasy
• Sci-Fi Adventure
• Sci-Fi Cult Classics
• Sci-Fi Dramas
• Sci-Fi Horror
• Sci-Fi Thrillers
• Supernatural Sci-Fi

But no "Dark Suspenseful Sci-Fi Horror Movies" or "British set in Europe Sci-Fi & Fantasy from the 1960s," as listed under the raw data in the Atlantic article. On one hand, I understand that by listing all altgenres and associated titles, you're showing your hand. But I'm sure they have enough smart people who could code a dynamic site to only show the genres that currently stream or ship content.
posted by filthy light thief at 10:49 AM on January 2, 2014

engineers, not ... user experience professionals, information architects or content strategists

Or, god forbid, film buffs. It doesn't seem like film, you know, the art form itself, is even a recognizable domain of knowledge to either Netflix or this Atlantic writer.
posted by RogerB at 10:53 AM on January 2, 2014 [5 favorites]

like film theory as reinvented by Engineer's Syndrome

understanding fiction is one of those lost AI dreams from the seventies, the protoblog robotwisdom goes on and on about this. its actually a hard problem that you can't just throw "machine learning" at.
posted by ennui.bz at 10:55 AM on January 2, 2014

It doesn't seem like film, you know, the art form itself, is even a recognizable domain of knowledge to either Netflix or this Atlantic writer.

I know! And the incredulous tone that they actually trained people to watch movies and describe them! Wow, what forward-thinkers. Have a pat on the back, Netflix.
posted by bleep at 10:56 AM on January 2, 2014

Somewhere, a Library of Congress-using librarian is snarling, 'Netfliiiiiiiiiiiix!' into their screen. Controlled vocabularies adept at fitting thousands of different genres and subjects and mainly covering one material type? Yep, we've got that.

Clearly the Atlantic writer had no clue that librarians do this, but I'm curious whether Yellin's group knew or took advantage of past classification misadventures. I hope to hell someone in the group read some Otlet or Briet or Ranganathan.
posted by librarylis at 11:22 AM on January 2, 2014 [12 favorites]

One of the many, many terrible things about Canadian Netflix is that no matter what you search for you get a maximum of 10 results and cannot access any more. Even if you just enter "the," you're still only going to get 10 hits and if it's not what you're looking for, well, too bad.

Canadian Netflix: It's A Good Thing It's Only Eight Bucks A Month
posted by The Card Cheat at 11:26 AM on January 2, 2014 [1 favorite]

Last night I was looking through "Cerebral Movies" and saw that they included The Human Centipede 2 so, you know, good job on that system, Netflix.
posted by jason_steakums at 11:29 AM on January 2, 2014 [10 favorites]

Isn't the point that you don't need film buffs to help quantify and classify films? I mean, using microtags to classify small bits of info, as well as when they occur in the movie/show is a pretty good idea if they want to create their own material. I don't think Netflix should attempt to see itself as a streaming powerhouse - I'm not sure how viable that business model is anyway. They still need to license movies from other places, and so this microtagging is a way to extract/create valuable info from/about their existing product. They can then leverage that information to create their own shows like House of Cards and Orange is the New Black - and to me, that's a more reliable profit center than licensing others' content.
posted by antonymous at 11:30 AM on January 2, 2014

Last night I was looking through "Cerebral Movies" and saw that they included The Human Centipede 2 so, you know, good job on that system, Netflix.

It's an extended metaphor/thematic exploration of Sherman's March to the Sea, you philistine.
posted by griphus at 11:38 AM on January 2, 2014 [4 favorites]

antonymous, I think the point is that they're collecting tons of information, but it sure seems like it's mostly the wrong information.
posted by straight at 11:39 AM on January 2, 2014

fings: "Playing around with the URL they give is fun, though"

In choosing random numbers in the URL between "0" and "76,897," it appears that some don't work, but it looks like they've got categories just waiting for movies that apparently don't exist yet in the Netflix inventory. Take note, Hollywood: if you're looking to develop Gay & Lesbian Mother Daughter Dramas, Netflix has a genre reserved just for you.
posted by Dr. Zira at 11:40 AM on January 2, 2014

"1337" gives me "Movies starring Kristine Hermosa" and it is empty.
posted by griphus at 11:42 AM on January 2, 2014

Oh, and hey, just in case you need to make an 8 to 10 year old weep you're all set, just as long as they are not Quirky Talking-Animal Children.
posted by Dr. Zira at 11:45 AM on January 2, 2014

Quirky Talking-Animal Children
Quirky-Talking Animal Children
Quirky Talking Animal-Children
posted by jason_steakums at 11:48 AM on January 2, 2014 [4 favorites]

Quirky-Talking Animal-Children
posted by zippy at 12:01 PM on January 2, 2014 [1 favorite]

Netflix's PR liaison, a Dutch guy who keeps a miniature windmill on his desk named Joris Evers

Is that the windmill's name, or the guy's? Or the desk's??

Also this is super-fascinating. Every time I open up Netflix I marvel at the bizarre list of categories.
posted by uncleozzy at 12:09 PM on January 2, 2014 [2 favorites]

I'm pretty sure it's the name of the desk. I've heard naming furniture is one of the Netflix enforced quirkiness schemes, in an attempt to get the engineering-minded folks to think outside the box.
posted by filthy light thief at 12:12 PM on January 2, 2014

They can then leverage that information to create their own shows like House of Cards and Orange is the New Black

HOC is a remake of a British show, and I thought OITNB was originally a novel or memoir?
posted by feckless fecal fear mongering at 12:15 PM on January 2, 2014

The Card Cheat: Canadian Netflix: It's A Good Thing It's Only Eight Bucks A Month

For your enjoyment: Netflix Canada Complete List dot blogspot dot com and Canadian iStreamGuide.

See also: Hola Better Internet browser add-on, if you browse via a computer.
posted by filthy light thief at 12:16 PM on January 2, 2014 [1 favorite]

They can then leverage that information to create their own shows like House of Cards and Orange is the New Black

feckless fecal fear mongering: HOC is a remake of a British show, and I thought OITNB was originally a novel or memoir?

Yes it is. I think the point is that Netflix picked to produce or back those two shows over other competing possibilities because they knew what people liked (or what people watched based on what Netflix thought people liked).
posted by filthy light thief at 12:17 PM on January 2, 2014

Yeah but their fancypants algorithm wouldn't have really added much value there. What is everyone watching? Longform dramas with high production values. Don't need to know "People who liked this movie also liked.." in order to see that. Sopranos, Six Feet Under, Mad Men, etc etc etc, the trend has been going for a while. HoC broke slightly new ground inasmuch as I don't think any of the new style of longform dramas really covered politics/Washington (West Wing, I think, was straddling oldschool episodic tv and new-style longform) in this way. And OITNB is kinda sorta Oz with women, no? Point being, anyone could have looked at those options and picked them to be at least moderately successful.
posted by feckless fecal fear mongering at 12:52 PM on January 2, 2014

I think their algorithm does add value and is heavily relied upon to influence the production that goes into shows. They know when you pause shows, as well as if/when you might stop watching a series entirely (and the events that preceded that decision) - that's data that can be used at a much more granular level than just saying, "let's produce some longform dramas."

The example of this that comes to mind actually happened in OITNB - I knew of Netflix's viewer metrics going into the show. And I think many people like myself (maybe just males in general) may be skeptical about a show that features almost nothing but interactions between women. But the first minute was basically a non-stop lesbian sex scene (or at least lots of naked women, I can't remember now), all but guaranteeing a rapt male audience. I don't think Netflix produces that show without years of gathering data on viewer habits (see also their decision to release everything at once for the binge-watching crowd).
posted by antonymous at 1:11 PM on January 2, 2014

I think the whole algorithmic approach to content generation is neat, but while it's being touted so much it's easy to ignore the real thing Netflix original series have going for them, which is simply that putting a show that would really risk becoming the typical "critical darling that struggled to find an audience on television" on a streaming video-on-demand service turns it into "critical darling that a shitload of people will get around to binge-watching eventually". Way more opportunity for anyone curious to check it out when they don't need to catch it live or fill up their Tivo or buy the DVDs later.
posted by jason_steakums at 1:15 PM on January 2, 2014

But the first minute was basically a non-stop lesbian sex scene (or at least lots of naked women, I can't remember now), all but guaranteeing a rapt male audience.

See also: The L-Word, Game of Thrones, and oh anything aimed at the male gaze. Putting hot naked women into a TV show to guarantee a rapt male audience isn't exactly an unknown strategy.

Nothing about HoC or OITNB couldn't have been done without this algorithm, as it's basically all good TV storytelling. They may have used it, sure, but given that HBO (etc) were making wildly successful shows long before this algorithm happened (and let's be honest, neither HOC nor OITNB, good though they may be, have received the same level of attention as e.g. Game of Thrones or The Tudors) and will continue to do so after, as was pointed out above, it's realized that this algorithm isn't as useful as it appears to be.
posted by feckless fecal fear mongering at 1:26 PM on January 2, 2014

Does this explain why I have a category called: "recommended to you because you watched null"?
posted by vespabelle at 1:35 PM on January 2, 2014

French Hit-Man Vampire Movies Set in Asia From the 1920s For Ages 2 to 4 Starring Donald Pleasence

Oh yeah, I've seen those. Pretty entertaining even if you're above the targetted age group.
posted by Foosnark at 2:11 PM on January 2, 2014 [3 favorites]

"Boss, you're not going to believe this, but according to our metrics, males in the 16-35 demo are consistently pausing and re-watching shows . . . well, there's a few things here . . . explosions, defenestration, . . . but here's the surprising one: scenes with female nudity!

"I know this is going to sound crazy, but I was thinking, what if we tried putting a few nude scenes in that new show idea we're kicking around? Maybe we could pick-up a bigger slice of that demo?"
posted by straight at 2:14 PM on January 2, 2014 [3 favorites]

I think many people like myself (maybe just males in general) may be skeptical about a show that features almost nothing but interactions between women. But the first minute was basically a non-stop lesbian sex scene (or at least lots of naked women, I can't remember now), all but guaranteeing a rapt male audience.

If this is the product of Netflix's algorithmic data wankery, then I for one could do without it. Also, wow - "a show about women talking would totally suck, but then it turned out they were naked so it was awesome!" is not a nice sentiment.
posted by Dysk at 4:19 PM on January 2, 2014 [5 favorites]

"Experimental Slashers Based on Bestsellers Set in Latin America From the 1920s For Ages 0 to 4"
posted by mittens at 4:23 PM on January 2, 2014

All my recommended categories make me sound unbearable. "Cerebral French movies of the 40s"? "Critically-acclaimed Visually-Striking Dramas"? "Sundance Film Festival Award-winning Dark Movies"? Step away from this person.
posted by acrasis at 4:26 PM on January 2, 2014 [2 favorites]

Does this explain why I have a category called: "recommended to you because you watched null"?

Oh, that's just how Netflix copes when your view history is full of Asylum movies.
posted by JHarris at 4:48 PM on January 2, 2014 [3 favorites]

Violent Comedies Set in the Victorian Era

Burke and Hare? Or I guess technically that's 'Georgian', as it was about a decade shy of Queen Victoria. Funny movie, tho.
posted by FatherDagon at 5:14 PM on January 2, 2014

Also, wow - "a show about women talking would totally suck, but then it turned out they were naked so it was awesome!" is not a nice sentiment.

A bunch of dudes are probably not going to get together and watch a bunch of "women talking" even if there's nudity, but Netflix (and everyone) knows that tons of TV time is spent by couples together (and four eyeballs is better than two). I mentioned the nudity at the very beginning of the show because it's a clever hook which actually didn't get exploited as often as I expected. In terms of film/TV as art, I think extended use of viewer data is a race to the bottom.
posted by antonymous at 5:53 PM on January 2, 2014

I'm just popping in to bitch about Netflix's categorization of any films not made in the US.

"
Because you liked Der Baader Meinhof Komplex, we recommend:

• Whore's Glory
• The Good, the Bad, and the Weird
• The Human Centipede
• Purarambo
• Dead Snow
• Sexual Chronicles of a French Family
• Tai Chi Hero
• Nosferatu

"

Really? Sigh.
posted by Seeba at 8:02 PM on January 2, 2014 [1 favorite]

That is to say - interesting read, but as detailed as the setup is, it doesn't seem to actually work very well.
posted by Seeba at 8:05 PM on January 2, 2014

Also, the system needs a wildcard (yes yes, I realize it's entirely subjective... ... well... mostly subjective) that allows users to request "with a minimum of plot holes" or "well written".

I don't particularly care about genre, but I appreciate movies with efficient use of language, or that have a written style deeper than "Hey look, a threat, let's blow it up!"
posted by Seeba at 8:09 PM on January 2, 2014 [1 favorite]

And then of course on the front page of Amazon Prime you will find "Dramas" and "Comedies".
posted by Brocktoon at 8:30 PM on January 2, 2014

What? I mean, even I could probably write such a script, given a bit of time. He should have been pointed to Python, which is free.

Don't know anything about the software he talks about, or the data he was collecting, but I can tell you one thing: Never promise to write code that will scrape anything off the web, except maybe static HTML produced by code written by anal-retentive fascists: that way lies madness.
posted by Dr Dracator at 5:48 AM on January 3, 2014

I mentioned the nudity at the very beginning of the show because it's a clever hook

A clever hook? Not it isn't. It's a variation on this theme at best.
posted by Dysk at 6:28 AM on January 3, 2014 [1 favorite]

I wish it would still look at my actual ratings rather than ignoring them to slavishly devote itself to "watched." My older set-top box doesn't support profiles, so Netflix keeps suggesting things based on random combinations of the tags from the very different tastes of 4 different family members. No one wants "Dark romantic comedies set in England," because "dark" is one person, "romantic" another, "comedy" a third, and "set in England" a fourth. Netflix suggestions around here are now entirely useless to anyone.
posted by tyllwin at 11:18 AM on January 3, 2014 [1 favorite]

Felix salmon, as always, has a great take:

Netflix, then, no longer wants to show me the things I want to watch, and it doesn’t even particularly want to show me the stuff I didn’t know I’d love. Instead, it just wants to feed me more and more and more of the same, drawing mainly from a library of second-tier movies and TV shows, and actually making it surprisingly hard to discover the highest-quality content. It’s a bit like what Pandora would be, if Pandora was severely constrained in the songs it could choose from.

posted by So You're Saying These Are Pants? at 7:45 AM on January 4, 2014 [1 favorite]

That makes sense. Netflix has no incentive to get their customers in the habit of consistently watching films they love. It's much more in their business model to cultivate customers who are in the habit of settling for stuff they sorta like.
posted by straight at 6:34 PM on January 4, 2014

It's also what I'm willing to pay for at that price.
posted by Brocktoon at 12:36 PM on January 14, 2014

Netflix recommendations lately have been so poor that I've taken to just typing random words into the search box. If they are discounting the ratings for their recommendations that explains a lot.
posted by Mitheral at 8:24 AM on January 16, 2014