Every librarian has been asked at one time or another to recommend a book based on something that a reader liked. You know the conversation: “I just read <book title> and loved it, what else do you have like that?” And since the Dewey Decimal System is notoriously weak for categorizing fiction, there are companies that sell books and software to help librarians answer that question. Netflix also has a process to make movie recommendations, categorizing films into thousands of “micro-genres” which can be almost comically specific. The first article linked below from The Atlantic describes how Netflix does that, while the other articles are reactions to the Atlantic article. It’s interesting to see how someone else handles the “what do I read/watch next” question.
- How Netflix reverse engineered Hollywood (The Atlantic/Alexis C. Madrigal) “Using large teams of people specially trained to watch movies, Netflix deconstructed Hollywood. They paid people to watch films and tag them with all kinds of metadata. This process is so sophisticated and precise that taggers receive a 36-page training document that teaches them how to rate movies on their sexually suggestive content, goriness, romance levels, and even narrative elements like plot conclusiveness. They capture dozens of different movie attributes. They even rate the moral status of characters.”
- 76,897 genres and 1 very human touch: How Netflix knows you best (The Daily Dot/Beejoli Shah) “Given its recent successful dips into original programming with House of Cards and Orange Is The New Black, Netflix may also be using its vast expanse of metadata to make its own programming even smarter. While metadata can’t create a show, it can give Netflix’s development executives … a distinct advantage into viewer preferences that focus groups just can’t.”
- Netflix’s dumbed-down algorithms (Reuters finance blog/Felix Salmon) “While Amazon has orders of magnitude more books than your local bookseller ever had, Netflix probably has fewer movies available for streaming than your local VHS rental store had decades ago. At least if you’re looking only in the ‘short head’ — the films everybody’s heard of and is talking about, and which comprise the majority of movie-viewing demand. So Netflix has been forced to attempt a distant second-best: scouring its own limited library for the films it thinks you’ll like, rather than simply looking for the specific movies which it knows (because you told it) that you definitely want to watch. This, from a consumer perspective, is not an improvement.”
- Marketing lessons from the Netflix’s micro-genre generator (Marketing Pilgrim/Cynthia Boris) “The takeaway here is that data is your friend. Dive in and see what’s really going on with the business then craft a campaign to either capitalize on your big sellers or push out that old merchandise that’s not moving. People will buy anything (Visually-striking Foreign Nostalgic Dramas) if you package it right.”
Burr fact:
For some reason that is a mystery to Netflix, their data places Perry Mason stars Raymond Burr and Barbara Hale among the top ten favorite actors/actresses.