Meet a King of Netflix

By Alexis C. Madrigal
NetflixGraphic20k.jpg

When we ran the story yesterday about Netflix users who've rated tens of thousands of movies, I was hoping that one or more of them would come out of the woodwork.

And so one has. Meet Brian Dear. He's writing a book about PLATO, which he describes as the first online community in the world. And he's rated 20,348. You can decide for yourself which is the more impressive accomplishment.

In any case, after he posted in the comments about how many movies he'd rated, I got in contact with him, asking for an explanation of his motivation and method. Turns out that one reason he did so much rating was as part of his project to evaluate the Netflix user experience earlier this decade. He documented all that. The idea was to get Netflix to change. So not only was he a power user of the rating system, he sought to change the company that built it, too.

Here's why and how he rated 20,348 movies in his own words -- and what he thinks about Netflix's algorithm after all that effort:

Netflix was relatively new back in 2002, and I felt there were lots of good things and a few not so good things about the user interface, and pointed them out and made suggestions for improvement. It wound up getting the attention of the Netflix execs and product team and we had some good exchanges and I felt "mission accomplished."

I rated movies I'd never seen to tell Netflix "no interest" -- in the hopes that if it knew what I did *not* like, as well as what I *did* like, it could only help in terms of recommendations. Another major motivation was that back in those days, one was not able to tell Netflix that one did not want to see **anything** in a particular genre. You can block entire genres now (thanks to my prodding, perhaps). So the only way back then -- so I thought -- to drill home the point to the Netflix recommendation engine that I didn't, say, want sports movies, or TV programs, was to say "not interested" to everything in the genre.

It didn't help, amazingly. I might say "not interested" to 500, 1000 movies or programs in a particular genre, and it would STILL recommend stuff from that genre, which only emboldened me more :-) And to my surprise, all that effort wound up breaking their recommendation engine! To date, they've never fixed it. Something I once pointed out to Reid Hastings, when I bumped into him at a conference. He laughed and told me forget it, they'll never fix it. And so, years go by and I keep renting occasionally from Netflix but I never get any recommendations from 'em.

I haven't rated many movies since 2005. Prolly a dozen or two. In other words, this isn't an ongoing preoccupation.

Take a look at the attached screen grab -- note that Netflix is broken. I've rated 20348 movies, but it cannot make any suggestions, and tells me to rate more "so we can help you find movies you'll love".

If you're a Netflix power user -- or you've just got an interesting algorithm training methodology -- do get in touch with me. I find you people fascinating.

This article available online at:

http://www.theatlantic.com/technology/archive/2010/09/meet-a-king-of-netflix/63191/