Viralogy

Personalized product recommendations can be done on any eCommerce website. As Darren Vengroff of Mashable points out, just about anyone can figure out how build an application that utilizes Facebook Social Graph to collect data for usage. He also points out, however, that not everyone knows how to leverage that data correctly - there is just too much information to be had.

The Problem for Personalization Tools

Say a merchant decides to build a personalization tool for his eCommerce website. In Vengroff’s article, there is an example of a teenage boy whose public Facebook profile shows that he “likes” the Porsche 911. It is pretty likely, considering the boy’s age, that recommending a Porsche to the boy on the front page will not lead to him buying it. This piece of data is useless in this sense. But, as Vengroff continues, the boy might potentially buy a tshirt with the image of a Porsche on it. How, then, does the merchant’s tool know when to correctly use the data that has been collected? Vengroff demonstrates the complexity of the situation with his paradoxical headers “If You “Like” It, You Might Want to Buy It” followed by “Liking Doesn’t Always Lead to Buying”; in other words, if customers “like” it, they probably want it, but, they won’t necessarily always buy it.

Traditional recommendation engines are based off of data such as previously bought items and items additionally bought by others (think Amazon). The new personalization tool is different because it gathers data from a dynamic structure whose primary aim is not for people to shop, but to connect with others within it. Merchants must realize that no matter how it may be utilized, Facebook is still a social network where people gather to interact. Information that is listed may or may not mean anything: sometimes “liked” items are purely aspirational, as Vengroff notes, or they may even be for reasons that are meant to be purely comical.

While for some pages, “Likes” indicate purchase propensity, there are many other pages for which “Likes” tend not to indicate purchase propensity.

What This Means

Context and relevance are vital to proper utilization of social commerce. Vengroff recommends that merchants continue to offer “a multiplicity of recommendations” in order to counteract possible challenges, but keep in mind that recommendations still have to be narrowed down in order for them to be useful. Tools and applications that attempt to personalize recommendations must then consistently be updated in order for the system to be able to grow and effectively weed out irrelevant information.