There are a number of ways to optimize an ad’s performance in social media by using psychographic insights and expanding on the semantic fields of the concepts that are associated with the ad.  In the next three blog entries, we'll look at a few of the shortfalls of many existing ad targeting systems and suggest some solutions:

1)       Potential Targets are Missed
It is very easy to miss potentially high-quality targets if the explicit and implicit interests and psychographic features of a user’s profile are not fully leveraged. For example, let’s say that you are trying to target an ad for Hiking Gear.  Wouldn’t you want to send an ad for Hiking Gear to someone who has written the following on their profile in a social network:
“I love roasting hot dogs over an open fire after a long day of exploring the great outdoors”  
Of course you would!  But if you are not using an ad targeting system that understands how interests are correlated, but rather have simply chosen a few keywords, you might easily have missed this high-quality target.  In such a case, which is actually quite common, the ad serving platform has to be able to uncover a variety of relationships amongst the user’s interests and the products or services being offered.

One of the central problems here is how close the relationship between the ad and the potential target profile ought to be. If a statistical model is being used by which weights are assigned to concepts in a vector space to describe how close they are to each other, then what would be considered close enough to warrant matching the ad?

For example, should an ad for Snowboarding Gear be shown to someone who says he likes Skiing?  Most word-clustering algorithms would probably assign a very high degree of closeness between Snowboarding and Skiing.  But while they are “close” to each other, do actual people in the real world who are interested in one have an interest in making any purchases associated with the other? Or do people really tend to participate in only one of these sports, rather than both?

There are many cases like this, and one way to overcome this kind of problem is to look at the person as a whole. Taking all of their interests and opinions into consideration may help to come to a better holistic understanding of their likes and dislikes and can help solve the problem presented above. So, for example, if someone says that they like Skiing, Winter Sports, Surfing and Tactics brand products, then we could in fact be pretty sure that they will also respond well to an ad for Snowboarding, but if someone says they like Skiing, Switzerland, Classical Music and Yachting, they probably are a more traditional skier and would not respond well to Snowboarding ads.