By now, if you are a follower of all things tech, you have probably heard the term “social graph.” The social graph is a collection of data points noting what you and your friends (Facebook and Twitter friends that is) like on the Internet.
The social graph is best represented by the ubiquitous “Facepile” – that Facebook widget that lets you know that Katie Perry, John Stamos and 300 of your friends also like something on whatever website you happen to be on.
The theory behind the social graph is that you are more likely to like things your friends like. Think of it as personalized recommendation from your neighbor – the web 2.0 version. So if your friend Bob likes a certain ice cream store, or better yet, a pediatrician, that ice cream store or pediatrician will create a personalized ad telling you that Bob likes them, in hopes that that little bit of info will make you more likely to click.
In case you were wondering, this is Facebook’s business model in a nutshell. Facebook creates value by allowing advertisers to tap into your social graph and deliver personalized, and therefore more compelling, advertisements.
This all sounds great, but what if you are friends with someone who actually has very different tastes than you? Let’s say that you are a die-hard Mets fan, while your sister goes to every Yankees home game. Ads that tell you that she loves the Yankees aren’t going to make you switch teams.
And therein lies the social graph’s Achilles heel. It’s not always safe to assume that you and your friends have shared sensibilities.
Enter the interest graph. The interest graph is a mechanism for creating a taste profile that matches you with the likes and recommendations not of your social network, but of people who have similar tastes.
What that means is a website will track its users’ likes and dislikes and create user interest profiles. If your behavior starts to resemble the behavior of another user, or set of users, the website will begin to send you recommendations based on what other people who are like you, like.
The most common example of the interest graph in practice is Amazon’s famous recommendation engine. Instead of telling you what your friends bought, Amazon tells you what people who have similar purchase histories to you have bought. Like Fifty Shades of Grey? Well then you’ll probably also like Fifty Shades Darker.
The interest graph is being used by tons of web companies, from online dating services to music listening web apps, in order to deliver the optimal, most enticing offer possible.
Random Cocktail Party Fact:
Back in 2006 Netflix decided that the algorithm that recommends movies and television shows to users, called Cinematch, was so valuable to their business that they offered a million dollars to the first team who could improve it by just 10%!
Although one team was able to improve the algorithm by a small percentage within 6 days of the contest beginning, it took 3 years before a team was able to improve the algorithm by 10%.