• by Steve Smith , Staff Writer @popeyesm, August 7, 2009

As we explored in our Grill the Vendors panel at last week's OMMA Behavioral, the "optimization layer" of the ad network value chain has come on strong this year as companies like Rubicon Project and AdMeld help publishers squeeze the most revenue from a depressed CPM market. These companies help publishers find in real time the most lucrative opportunity among their ad networks in order to maximize the value of their remnant inventory.

In the next few months, however, we are likely to see more companies emerge that optimize on the buy side to ensure campaigns are executing more efficiently. One company that is trying to prove out the concept of buy-side optimization, Aggregate Knowledge, is a veteran of Web site content optimization for companies like Washington Post and Sam's Club. Its newly announced "Audience Discovery Engine" claims to map every ad impression against its own proprietary data as well as interesting combinations of third-party data to ensure that campaigns are reaching only their target audience. We asked co-founder and CEO Paul Martino to walk us through an early case study of the approach and the novel types of data that can be pulled together to locate audiences.

Traditionally, Aggregate Knowledge has been used by publishers to deliver targeted content experiences or product offerings to users based on clickstream histories and aggregated data among like-minded users. It tries to expand the retargeting approach of basic behavioral targeting by looking at a broader set of characteristics shared among people performing a specific behavior and so expand the addressable audience. In the last 18 months, the company has been rethinking this sell-side technology as a buy-side tool for agencies.

Aggregate Knowledge ran a campaign for a Fortune 500 company that was supposed to target executive decision-makers in U.S. companies with more than 500 employees. In an initial audit of the campaign, 37% of the ads delivered were off target. "They had no way of figuring out who were the decision-makers at Fortune 500 companies," says Martino.

In addition to its own proprietary algorithms and data, Aggregate Knowledge had the agency data, which let it audit every ad that was delivered, and it partnered with third-party data. "We were able to get data from guys like Hoovers, for example, so we could do IP matching of addresses to determine the probability that the people were inside those Fortune 500 companies," says Martino.

Aggregate Knowledge also partnered with companies like Digital Element (IP and geotargeting) and LucidMedia (contextual ad targeting). "We are running this inference engine on the spot and in real time," says Martino. "We are able to come up with a model which can say that here are the five characteristics most likely associated with the person in the Fortune 500 company. We could literally score people. We can say with 100% certainty or 90% certainty that a person is inside your target, and if they weren't in a certain probability we kicked them out. We used five or six data sources."

As a result of applying these multiple data layers as filters, Aggregate Knowledge claims its optimized campaign showed a 162% increase in the number of on-target impressions and a 48% increase in clicks over the unoptimized run. The company charges a percent of media spend. In the test case of the Fortune 500 campaign, Martino says the optimization achieved a 44% better cost per click and saved the client 60 cents on the dollar. "They paid us a percent on media that was well less than 60%, so they were pleased."

Of course buy-side optimization also adds yet another layer to a stack that some media buyers at OMMA Behavioral said was getting confusing and unwieldy. Ad networks increasingly tell me that education is becoming job one, because ad exchanges, trading desks, data exchanges, optimizers, etc. are creating confusion on the buy side that can retard the progress of this new era of sophisticated targeting.

"I think this is one of the most important things anyone in this space needs to care about," Martino agrees. "We are convinced that the eco-system simply is too complicated. It creates significant hesitation from the buyer to make decisions."

Some buyers prioritize simplicity and would just as soon have a third party bundle together solutions to insulate them from the confusion. Others want best-of-breed throughout the value chain and are willing to dive into the morass of choices and layers in order to get it. Martino suggests that the industry start offering "stack based solutions" that do the grunt work for the buyer and present them with a bundle that simplifies and delivers on quality. "We think absolutely we have to start stack-based selling where two or three of the vendors put together all the plugs so we can make it easy to buy best of breed stacks." He says the enterprise software industry has been pre-packaging partnered product this way for years and the ad industry needs to follow suit. "I believe 2010 is the year when these technologies are going to scale," he says. "And if the vendor ecosystem doesn't get wise, there is going to be paralysis in 2010 instead of scale."