Amazon is MicroTargeting for Your Shopping Cart
One of the most common (and powerful) features of consumer and social websites are their recommendations – whether it’s a new person to friend on Facebook or a new movie to rent from Netflix, it seems websites are always suggesting something new to us.
When it comes to making effective recommendations based on your online behavior, Amazon is the granddaddy of them all. Its recommendation method took over ten years to craft and perfect, but its superiority generates countless purchases from customers who’ve been swayed by the tailor-made suggestions at the bottom of the page.
One of the Amazon’s techniques is the social recommendation, the proposal of new items based on the behavior of users similar to the individual customer. So if Phil and Phyllis buy the same lime green beanbag chair, Macroeconomics textbook, and Bob Marley poster, Amazon might classify them as similar users. Then, when Phil buys the new Strokes album, Amazon might suggest that Phyllis purchase it as well.
Envisioned in this light, Amazon’s recommendation system is quite similar to TargetPoint’s Microtargeting method. We gather large amounts of individual level data, form groups of these individuals based on their similarities, and then use the characteristics and political beliefs of each person to predict the political beliefs of hundreds (or even thousands) of individuals with similar characteristics.
So if Phil and Phyllis both drive Honda Civics, make about $40,000 a year, and shop at thrift stores, we might conclude that Phil and Phyllis probably have similar political beliefs. Therefore, if Phil’s survey results show that he supports a certain candidate’s stance on immigration or health care, we can assume that Phyllis probably feels the same way.
Now consider this on a much larger scale, so that the number of people in a group is in the thousands and the number of characteristics upon which these groups are formed is in the millions. With such large amounts of data, the accuracy of these predictions becomes astounding.
Want to learn more about the Amazon recommendation system? Check out this article: Amazon.com Recommendations: Item-to-Item Collaborative Filtering (PDF)
Comments on this entry