The algorithm Netflix uses is easy to understand if hard to copy. Part of the reason is their ‘mounds’ of viewing hours logged which is hard to replicate – this speaks to the ‘quality’ of their sample.
The nature of Netflix’s algorithm is like so. When people say they like a certain genre, but actually watch something else, Netflix knows it. You know how people will say something to impress a date, well it’s the same deal.
“MANY PEOPLE TELL US THEY WATCH FOREIGN MOVIES AND DOCUMENTARIES, BUT IN PRACTICE, THAT DOESN’T HAPPEN.”(Wired)
How can you head Netflix off? What is popular at the box-office is still what audiences really actually watch. Yet they watch it so often it’s probably taken for granted versus the one art-house film they ever saw.
When Netflix says the meta-data of a film, they mean the ‘zeitgeist’ of a movie. But they go on to say, it includes the ‘sign of the times’ out of which a movie was made – kind of like Zelig. So it isn’t just film-school deconstruction, but also a point of view regarding stuff. If you like divorces that end well, you’ll see more of that and perhaps all sorts of near-misses that end well too.
“By looking at the metadata, you can find all kinds of similarities between shows. Were they created at roughly the same time? Do they tend to get the same ratings?”(Wired)
Other things Netflix uses to recommend a ‘new’ film to its viewers include Netflix member reviews. I do not think it relies much on outside sources like movie-critics and media exposure. Netflix is interactive in that way but the diversity of its subscriber base has it buying and creating different types of shows – of course the movie that appeals to the broadest taste will get produced exclusively for Netflix. That is the ‘interactivity’ of Netflix.