E-Commerce & The Social Graph

The social graph . . . the silver bullet for all things previously haven’t worked online. Seems like for the last 3 month, everyone has been recycling dot-bomb ideas, adding a “social graph” angle . . . and calling it the next great thing. Here is the deal. . social graph will ultimately be an important part of the formula for building the next great web application BUT its not a panacea for an idea that never worked in the first place. Understanding user behavior and context will still be the fundamental first step to creating the next killer app.

Buried in the various posts of the week, is this little gem on Venture Beat on eBay’s experiments with merging the social graph and e-commerce that anyone (entrepreneurs, pm’s etc) experimenting with the social graph need to take note.

What eBay has discovered. . . (and BTW Facebook learned the same thing too from Beacon) is that social recommendation does not spur an impulse purchase from another user. Extrapolating even more broadly, social recommendations do not cause a passive user to become an active user. One area where this data point could help predict success is in the display advertising targeting technology. Social graph based display advertising targeting companies (too many to mention) will not have the success that many think they will . . . display is still a passive medium . . . it will never be like search an active medium.

But is it true that “people want to keep social and commerce activities separate”? That is part I think eBay had it wrong. Its too broad of a hypothesis.

I believe where social graph will have the largest impact is when users are already taking an pro-active stance (for example, all kinds of “search” like activities) and the social graph based recommendations and applets can act as a catalyst to improve conversion. For example, Amazon’s recommendation system generates between 10-20% of its e-commerce revenue (based on common research analyst estimates) so in that context, when an user is already take a proactive action, recommendations can be very valuable. Social Graph applications is just another method for segmenting like-minded users . . . collaborative filtering gets to the same end just through a different mean (ie from actual purchase behavior) . . . so anything that improves the targeting and relevancy of recommendation would in theory improve conversion (and prove valuable) to the end user.

I, as a user, don’t necessarily need to know that Joe, my best friend, just bought a pair of Ted Bakers 2 seconds ago. What I do care about is when I eventually need to buy a pair of shoes, you remember to show me a couple pairs of Ted Bakers that I might like. Simple, certainly not ground breaking (. . . hell, collaborative filtering could already be far more relevant than social graph based recommendations . . . ) but doesn’t mean social graph could not be a small but important part of the inevitable march towards the next generation of web applications.