In Netflix - a Different Approach, I began the discussion of how Netflix has solved the video rental shop dilemma - huge demand for the newest hit releases, and low demand for thousands of other videos in each store’s inventory. By using a recommendation system skewed to older titles, they can slow demand for the hottest titles in those critical early weeks. While customers can still request those titles, they are unlikely to see them promoted at Netflix, a direct contrast to what one finds in the corner video rental shop.
This strategy also reflects the way online rental firms make money compared to brick & mortar video stores. Your neighborhood Blockbuster makes money when you rent a movie, and makes no money if you don’t rent a movie. Hence, they have great motivation to piggyback on new release advertising by the studios. They know the pre-release advertising for DVDs will bring customers into the store, and by combining in-store signage with lots of inventory, they can drive rentals upward. Netflix, by comparison, is paid by monthly subscription. They don’t make more money if a customer rents a movie; in fact, every time a customer returns a viewed movie for a new one, it costs Netflix money. Of course, Netflix doesn’t want to discourage customers from viewing movies - a customer who feels he’s not getting his money’s worth from the service is likely to cancel. Nevertheless, Netflix has no motivation to push incremental revenues of the latest hits. That will bloat their inventory in the short run, and leave them with excess inventory in future. There’s no doubt in my mind that if Netflix put the latest hot releases on each customer’s start page, they would multiply the immediate demand for those titles by many multiples; doing so, though, wouldn’t make Netflix a penny more in revenue.
So, should I, as a Netflix customer, feel manipulated because they keep the newest hits if not hidden, than at least out of prominent view? Perhaps a bit… but any annoyance with that practice is offset by their growing ability to find movies I actually might like but have never seen. Instead of blasting ads at me for The Dukes of Hazzard, the DVD currently being heavily promoted in television commercials, they are offering me Rear Window and Dr. Strangelove. They’ll even explain why they suggest these films. For example, I had previously positively rated The Graduate and Chinatown; apparently, others who liked these two movies also liked Rear Window. (I’ve seen it, and though it may not be Hitchcock’s finest, it’s not a bad recommendation.)
It’s clear this is a community-based recommendation. (Or, it’s a random choice to rent a slow moving title. 
) It’s highly doubtful that if you put Roger Ebert, David Denby, and a dozen other critics in a room that any one of them would say, “You liked Chinatown? And The Graduate? You’ve got to see Rear Window.” This recommendation could only be derived statistically.
If I have any quibbles with Netflix’s implemenation of this form of collaborative filtering, it is that the algorithm seems to favor popular titles. Hence, I seem to end up with recommendations for The Godfather Part II and Lawrence of Arabia. I’d prefer Netflix to use its community power to dig out some more obscure films that I might like rather than suggesting films that any regular movie fan is likely to have seen before. I’ve been tempted to build a profile based on obscure movies of the 1930s to see how good Netflix is at finding similar films, or if the come up with more off-the-wall recommendations of popular films from recent years. Using unadulterated community likes and dislikes is both a curse and a blessing. While it may come up with recommendations that don’t seem to make much sense based on genre, time period, etc., it also may come up with non-obvious choices that the customer will really enjoy.
What’s next? I’d like to see customer algorithm tweakability at Netflix, the equivalent of Google’s infamous knobs and dials. It would be nice to be able to change the weighting of various factors to trigger recommendations that are still based on community data but overlayed with my own preferences at that particular moment. Even with the current imperfect system, though, Netflix is clearly following the Web 2.0 roadmap of using a community of customers to greatly enhance the content and function of its web site.
Add this post to: del.icio.us - Digg it - Stumble it - Furl - Yahoo MyWeb No Comments so far
Leave a comment
Leave a comment
Line and paragraph breaks automatic, e-mail address never displayed, HTML allowed:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>