My Million-Dollar Plan
HENRY P. RALEIGH
NOW YOU DON’T get many chances to make an easy million dollars and I think I’ve got a good shot at this one. It’s a contest, you see, set up by Netflix, the Web-based DVD rental company, back in 2006 and yet without a winner although plenty have tried and are still working on it. You have to be a subscriber, of course, but for $16.99 a month Netflix will mail to you the number of films you select from its over 100,000 titles — postage paid return as well. Curmudgeon that I am, I figure 99,000 of these titles are most likely losers from the past and considering that four to six-hundred films are produced yearly and most of these are buried even before they hit the graveyard of DVD distribution — well sir, that doesn’t leave you much to choose each month, does it? And who is going to search through 100,000 titles to find something they might recognize, I ask you?
Right there you have Netflix’s problem. Customers sign up for a few months, run through their favorite films, gamble unhappily on some unknowns and then throw in the towel. The company’s tracking engine, Cinematch, does its digital best to second-guess what each customer likes or dislikes based on available viewer’s ratings and make helpful recommendations for further orders — with about 60% success. It’s the missing 40% that hurts. Suppose you had gobbled up Woody Allen films like “Take the Money and Run” and “Bananas”. Cinematch, insidiously recording this (personally I don’t want my taste in films inscribed in a data bank for history to snigger at), smugly tries to pile on Adam Sandler films — that’s enough to do you in, I think. No, Cinematch needs a better way to zero in on what you really want in the way of satisfying films without prying into your ancestry, medical records, or neighbors’ gossip. Here’s where the contest comes in. If you can come up with a scheme to boost the odds of predictability by 10% you win the million. The company reckons that that much of a jump is worth a lot more than a million in profits.
Up till now its been mostly mathematicians who have been puzzling over the data gathered up from a half-million Netflix customers who have scored 18,000 films with plenty more to go. These math types keep fiddling around with something called algorithms, which I’m sure, is a big mistake. I don’t actually know what an algorithm is; I didn’t in high school either, which is why I decided to become a cineaste — it’s common knowledge that cineastes have no use for algorithms. They also toss about mind-boggling terms like ‘collaborative filtering’ and ‘singular value decomposition ‘ which, to put it plainly, means nothing more than finding a sure-fire pattern of what each and every Netflix customer will order and keep those $16.99’s rolling in.
So far 9.94% of the desired 10% has been reached by the leading contestant — pretty good but short of grabbing off a million. The other top contenders are nearly as close, churning out algorithms a mile a minute but all have hit a mysterious point beyond which it seems impossible to go. It’s called the “Napoleon Dynamite” factor and it has them dumbfounded, a film that polls all over the place, from lowest score to highest and never settles into a reasonably predictable position. I know just how this works. I couldn’t get through my first viewing of “Napoleon Dynamite” — damn, another teenage high school flic, nerdy, goofy kid beset by bullies — these clunkers are a dime-a-dozen. Endlessly making its way through cable film channels, I watched a bit more a second time, tried again — hey, not bad, quirky, some original stuff in there all right — maybe I even liked it, albeit reluctantly. There’s the rub, you see, films like “Napoleon Dynamite”, usually indies, produce such an erratic, capricious reaction among film renters that the algorithms throw up their digits in despair. But there is a solution and I have it.
To begin with, I discard all algorithmic approaches to the problem and propose instead — now get this — the Ciné Alienation Mentale rule or, as I like to say, the CAM (I use the French, more or less, to keep up that fine tradition in film-speak like montage and mise-en-scene). This takes several obvious observations into account and it’s surprising those algorithm fanatics never thought of it. Let’s begin with the basics:
1. People who subscribe to film rental services are most likely hooked on film, and may even be under the illusion their critical judgments are informed and worthy.
2. Such people, unfortunately, have only a tenuous relationship to reality and unknowingly suffer CAM. To put this another way, they aren’t playing with a full deck, yet are harmless in all other respects.
3. Through no fault of their own, the judgments of these mentally disadvantaged film lovers are about as reliable as you are of being absolutely sure to draw to an inside straight.
4. The above applies equally to amateur and professional film goers alike. In any given year, it is a fact that working film critics across the country agree on the best and the worst of the year’s films less that 60% of the time — what can you expect of the average film goer, I ask you?
5. The smarter anyone gets about film the more erratic to the point of total lunacy his or her assessments become. A thoroughly trained post modern deconstructivist can find enough significantly deep meanings in “Teenagers From Outer Space” to make your head swim.
So my proposal to Netflix is to go for the CAM rule. Cinematch looks over the renter’s previous picks, decides his next selections will make no rational sense whatsoever and recommend accordingly.
I will wait for my winner’s check in the mail, thank you.