360i’s David Berkowitz always has something interesting to say, and his recent MediaPost column is no exception. In it, he discusses the concept of social graph targeting – the ability to target consumers based on associations with each other, presumably within social networks – as a next gen targeting mechanism.
Certainly the ability to understand the expressed desires of a consumer, then gain visibility into his or her relationship network, is powerful stuff. But a couple questions spring to mind:
- On social nets, how are the two (relationships + common interests) consistently connected? I friend people for a variety of reasons – friends, co-workers, college buddies, and now, weirdly enough, parents – yet us sharing hobbies or a taste in movies is not a necessary criteria, or even factor in our relationship.
- How is the context of a user’s relationship in a social network environment determined? Is it reliant on user input (for example I explicitly state I am someone’s friend/co-worker/cousin?)
- Do we have any more insight into a user’s frame of mind when in a social net environment? Data from last year indicated soc nets users were less receptive to ads and made fewer purchases as a result. Does new data tell a different story? How does this info help shape expectations around what soc nets can and can’t do from a marketing perspective?
The deliberate connections web users create with one another are significant, but understanding the commonalities that bind them makes it so much more meaningful. Am I connected to you because you give great advice on introducing baby to solid foods? Or because you’re a relative of mine? Different scenarios have vastly different marketing implications. The idea of peer-to-peer connections isn’t unique to social networks of course. Increasingly, web content is becoming socially-charged, allowing marketers to see how relationships are formed around topical content — i.e. which sites link together around the movie Bruno — which can help make targeting even more powerful.


Follow Us on Twitter
Join Us on Facebook
Nick Lim wrote,
Hi Valerie, interesting commentary.
I think the value of the socialgraph in ad targeting will be less about showing ads that my friends have seen/responded to, than about how does the socialgraph actually improve the predictions used today in ad targeting.
Most CPC ad servers have some sort of predictive function that says how likely person A is to click this ad. It looks something like this: likelihood to click for person A =demographics(A) + clickstream(A) + context + keyword (greatly simplified)
So socialgraph should just add to this like so
likelihood to click for person A = demographics(A) + clickstream(A) + context + keyword + socialgraph variables.
This is where it gets fun.
Nick
| Link | March 3rd, 2010 at 4:39 AM
Valerie Combs wrote,
Nick -
Couldn’t agree more re: the prediction piece. Thanks so much for the insightful comment!
valerie
| Link | March 5th, 2010 at 11:29 PM