10% better, 10,000% more effort required

While online marketing opens up a lot of different and exciting opportunities, it also offers many challenges. One of the biggest issues is the disparity between effort or ability and actual payoff.


This is a problem that occurs normally at startup, where the majority of online business (dotcoms, or blogs, etc) tend to be ignored, and only a few ever hit it big, but what I actually want to talk about is the case of The Netflix Prize

Have you heard of Sisyphus? He is a character out of Greek mythology that pushes a boulder endlessly up a mountain. This illustrates the problem faced by the Netflix recommendation system.


Netflix is an online video rental site, where customers order DVD's online, then receive it via mail. Since there is no physical location, Netflix can make use of The Long Tail, (more info on that here), but at the same time, it leads to an interesting problem, (discussed in the third point in Wired's article): How do you get customers to order the less popular movies?

Netflix came up to the same solution that most sites come up with - a referral system. By figuring out what kinds of movies customers would likely rent based on previous rentals, Netflix can lend out more discs, (which does not actually improve their income on its own, but makes it much more likely that their customers will not cancel the service).

Unfortunately, beyond a certain point, it is far to difficult to predict what movie a person would rent - people are almost random on which movies they like or dislike at times. Since it is online, people are also wary about revealing identity, so Netflix tries to be as unobtrusive as possible, using only past rental history to base predictions. At this point, Netflix came up with a brilliant idea: Ask the customers to create a better recommendation system - Wikinomics at its best!


The Netflix contest provides this opportunity - first person to improve the recommendation system by a mere 10% would win a million dollars! But much like Sisyphus and the mythical boulder, beyond a particular point (seemingly around 8%), any improvement becomes almost impossible. The million dollars would be a minor cost to Netflix, considering that a 10% increase in recommendation efficiency would be worth much more - the whole thought and concept behind it is long, so I won't go into it further, but you can read about it yourself here.

The point of this post is mainly to highlight the brilliance behind the contest (improve the company, create publicity, and create a user community built around sharing of programming information), and the inherent flaw of the Long Tail (difficulty of access). And maybe to talk about Sisyphus a bit. That guy deserves some PR buzz - his job sucks.