Once a company defines its best customers, it can use their "social graphs" to find others just as likely to buy the product. OK, but whenever we hear that term "social graph," our eyes glaze over. The former general manager of ad strategy and product management for Microsoft wants us to wake up to its power.
Behavioral Insider: What does this "social graph" notion mean to Media6 and its model?
Doran: Every social media or UGC site has a social graph of people who are actually connected. The Facebooks and MySpaces look at the little vertical social graph of people who are on their site, based on the explicit friends, not their communications. That is a common way of thinking about connectedness. From the Media6 standpoint, it looks more like pure network theory. It is not about the people that are actually listed but about people who communicate within networks. There are people who only communicate within Facbebook and connections that happen within MySpace and photos in Flickr. Things happen between these networks, and you end up connected with people in a very horizontal way instead of vertical. We take a more horizontal look at it.
BI: A lot of ad targeting solutions are interested in the nature of the conversation, but you are most interested in the connection.
Doran: The technology of behavioral targeting systems actually work when you look at what is being shared, and where people are going. It is very rich data. But we believe that the real power and the scalability of this social graph area lay in the connections between people. The power is in the homophily, or the connectedness.
BI: What does that like-mindedness get you as a marketer?
Doran: People really do communicate with those who are similar to them. Most traditional marketing mechanisms are dilutive in nature. You start with a marketer served with rich segmentation and rich understanding of who their customer is -- say, sophisticated males, interested in a well-branded product, appreciate collegiate sports, and play on the weekend. We dilute that down to men 18-to-35 with income greater than R45,000 and live in certain Zip codes. Isn't it better to start with that known entity of that user? If we can find the people that are connected with that person, we create a tight affinity cluster or group that would share a rich homophily or commonality of traits. The premise: we build custom segments, customs audiences for the marketer based upon the social graph. The advertiser tells us who their best customers are and we go find what we call the network neighbors that surround those users.
BI: What is the method of data collection?
Doran: We find these users across the social media and UGC sites. We wanted to go against data that was easily accessible where the rights of the data were already stated. We focused on the data the advertisers get in the process of serving ads on these sites. We participate with ad servers and networks. We have the same information that the advertiser had been getting and put rich algorithms on top of that to mine the connectedness that has been observed but never utilized.
BI: What kind of information in the ad-serving back channel would reveal connectedness to others?
Doran: Two major components are the ability to look at the URL where the ad was served and the ability to look at the cookie of the user looking at that profile page. In social media, there is a common taxonomy of how social media sets up the URLs. Within that, you see anonymous persistent unique identifiers within the URL in 95% of social IDs, blogger IDs, author ID numbers. That is how we are able to collect data as a proxy of connections between individuals without knowing the nature of that connection.
BI: So how does this translate into a targeting strategy?
Doran: It is the story of re-marketing. There is no dilution in that customer segment, and you get great response rates. But the problem with re-marketing is that it just doesn't scale; it is hard to find those users again back on the Internet. We say tell us where your best customers are and put a Media6 pixel on that site where you find those great users. Not only will we go back and market to that best customer, we will find the network neighbors connected to that person. For every person they give us, we usually find 10 people that are connected to that user. So we are building an audience based on data from the advertiser. We go out and find those people not only on the social-media sites but on the content-based sites the advertisers want to be in, whether it is article level pages or prominent spaces where media tends to perform a little better.
BI: Can you give an example of this at work in a promotion?
Doran: If a marketer has a high-quality brand site with low traffic, like 300,000 uniques, we will monitor their home page. I can usually see multiples of that by 10 and find 3 million uniques we can go out and market to bring that traffic back. Larger-scale, broader entities where people come in the front door look like everybody else;,we tend to go down to the product level information. People that are clear intenders to a product or service or who have bought product. We just expand that, and it that can be five to tenfold because those are the people connected to them.
BI: Do certain marketing objectives and product types work better with this approach?
Doran: This service works incredibly well for well-defined brands that are tribal in nature. If one person buys a product and other people see it, they actually go and communicate the benefits. We help accelerate that propagation. The tighter and more well-defined the brand is, the tighter the network neighbor connections around it. It performs really well for a direct response or branded message because you are coming down to a qualified audience. The places it probably doesn't work as well is when you are pushing commodity-based services that everybody basically has and has limited differentiation and multiple choice.