Social Media : The 
Social Graph Breaks Free

New techniques for mining social data for relevance and results

But what to do with such data once it's been extracted? Mostly, it's been used to create behavioral targeting segments for use in advertising campaigns that run on the same social network it was gleaned from. Some use social graph data to analyze the flood of conversation and identify the most influential users for a given advertiser to reach out to with a social media marketing campaign.

There is a much more valuable and disruptive use for social graph data. It can—and soon will—be used to optimize the buying of millions of individual ad impressions on the newly minted, real-time ad exchanges.

The Rise of Real-Time 
Ad Exchanges
The exchanges are relatively new, but they are quickly becoming an acknowledged threat to the reigning kings of online advertising, the ad networks. Ad networks buy inventory from publishers, aggregate it, group it into content categories and then resell it at scale with a markup as fast as possible before the inventory they bought spoils.

Now, ad exchanges like Google's Ad Exchange 2.0 (released in September), Yahoo's Right Media, Microsoft's AdECN and more plug publishers directly in to a market in which they can offer their entire inventories for auction in real time. When a publisher is plugged in to the ad exchange, each time a visitor goes to the publisher's Web site, advertisers bid for the opportunity to show an ad to that individual visitor.

Each auction happens in just 50 milliseconds. First, the ad exchange offers information about the ad impression being bid on, such as site, category and viewer characteristics. Next, advertisers estimate the value of that particular impression. Then all the participating advertisers bid, someone wins and the winner's ad appears on the publisher site for that particular visitor. This is repeated for each ad impression a publisher serves.

This new structure for buying and selling online advertising bypasses the ad networks, allowing advertisers direct access to publisher impressions. It also trumps ad networks' publisher-level targeting with audience-level targeting, while also enabling powerful new automated targeting optimization strategies. In addition, it allows advertisers to skip the markup imposed by ad networks. Finally, because impressions are purchased in real time, exchanges avoid the conflict of interest inherent in the ad network model of buying and reselling publisher inventory before it spoils, while promising to optimize prices and returns for both buyers and sellers.

Leveraging Social 
Graph Data in 
Real-Time Bidding
The trick to buying effectively on ad exchanges is calculating how much each impression is worth to a given advertiser—and that's where social graph data first becomes incredibly valuable.

During the evaluation phase of the real-time bidding process, social graph data can be used to buy impressions targeted to a particular audience with unprecedented accuracy. If you're sure who your target audience is, and you have good social data keyed to cookies, you can check whether each impression you bid on falls squarely into your target before you decide to buy, drastically decreasing media waste.

Now, consider the optimization possibilities. As an advertiser, you buy an impression on a publisher site using an exchange. That visitor clicks through to your site and converts, purchases or subscribes—whatever your desired action is. During a later auction, you check your social graph data and find out you're bidding on a close friend of the person who converted for you earlier.

How much more might that friend be worth to you? Demographic patterns tend to run in groups—many of your friends will be geographically, demographically and socially aligned with you. You almost certainly share interests in common, frequently live near each other, and probably discuss and recommend similar kinds of entertainment. Even better, there is a much greater chance that the customer who converted has talked about or sampled your product with the prospect, priming him ahead of 
your impression.

Of course, social graph data isn't the only valuable data source to use for optimizing online media buying; behavioral, geographic and internal customer databases work very well, too. But social graph data promises to be a particularly effective way of predicting which impressions to buy and how much to pay for them.

By using powerful sources of data like social graphs with predictive modeling, you will uncover key indicators of likelihood to buy that will help you buy more and more efficiently. The more data about your audiences you integrate, the more you can learn about the drivers for your online conversions. In this way, exchange buying with data becomes a huge focus group that can uncover demographic, behavioral and social triggers for purchasing your products, enabling you to constantly learn about your audience and then immediately put those insights to work in your media buying strategy.

Some Issues Remain
While this sounds great, there still are 
hurdles to clear. Historically, the number of cookies available via third-party 
social graph data providers has been low, 
but the numbers are rising quickly.

Network effects are stronger for things that incite buzz—in other words, if what you're advertising is going to spur people to tell others about it, it works especially well. It helps to focus on products where the network effect is strongest.

So even if you're not able to penetrate the conversation flow on Twitter or Facebook to reach enough of your users, you still can take advantage of the potential of social graph data. It can help you more precisely value impressions in real-time bidding systems and also continuously refine your audience modeling and segmentation for better results going forward. While it does not negate social media marketing—relationships with brand enthusiasts won't stop being valuable—it does have the potential to make your media buys much more effective and efficient. It might even help you learn things about members of your audience that they didn't know themselves.

Israel Mirsky is director of marketing at [x+1], a predictive marketing solutions firm based in New York. He can be reached at imirsky@xplusone.com.

• MORE ON
• SEARCH & SOCIAL MEDIA