One person who has strong views about the way ad networking environments are evolving is Rich Frankel, president of Rocket Fuel, the ad technology platform that combines multiple targeting approaches. Frankel was part of the team that built Yahoo's behavioral targeting network, so he has been around the block and seen the ad technology stack develop in recent years. For Frankel, the process of planning and executing online is perhaps too complicated for humans. He is a big believer in the role algorithms and automation will play in digital advertising's future.

 "It is very, very unstructured, messy, hard to understand and very rapidly changing," Frankel says of current ad technologies. He warned me when we started the conversation that he would drop the term "value" a lot because that is where he thinks things will shake out in this proliferation of networks, exchanges, and data providers now pouring into the system. "The value side is, who is really going to create value and who is going to enable commoditization?"

Frankel says that the ad exchanges, in trading ad inventory and selling to the highest bidder, are looking for efficiency and commoditization. "A huge challenge for the Internet for 2009 is understanding the value of the commodity," he says. Frankel argues that the exchanges may be adding some value in marginally increasing CPMs for publishers and helping the buyer bid efficiently, but the core problem is in discovering the actual value of the impression that is coursing through the system. "The buyers have no idea what they are bidding on," he claims. Are all impressions coming off of a Yahoo sports page really worth the same, he asks? "The exchanges liken themselves to stock markets, but they are stock markets with no transparency of quality, which is kind of shocking."

The new generation of data providers are now adding their own similar layer to the system with a range of new data points, from behavioral, to search to shopping, to social affinities. But Frankel argues that here too a blind marketplace has been created where value is determined mainly by how much people will pay. "But no one knows how good the data is. It is a random fact about a consumer. It is a category called 'men's shoes.' Do you want to buy some cookies in that category? We aren't telling you where the data comes from or how good it is."

Not surprisingly, Frankel's Rocket Fuel positions itself as sitting atop this stack, using the ad exchanges and the data providers as inputs. The layers beneath him, the data providers and exchanges, he views more as "the plumbing."

"The point of the business is to tie all of these things together and use it in a way to determine value for advertisers," he says. "We help them reach the right audience and deliver their campaigns effectively against the metrics, whatever they are." The platform uses multiple data points to try to determine on the fly what is the right impression to serve to the right person at the right time. "How do we understand the value of the data and the inventory holistically?" Frankel asks.

For Frankel, who ironically has his Princeton degrees in English and Classics, the future of online advertising lies in the machine and the math. "Our business model from Day One was to build a machine that can be better, faster, stronger at figuring out how to understand what consumers are interested in and what they will respond to." Rocket Fuel started executing campaigns in late spring, and Frankel says the system seems to work best with major brands. He says Rocket Fuel can touch about 95% of the U.S. Internet universe through a blend of exchanges and direct relationships with publishers.

One of the main things Frankel has learned since his days developing the behavioral engine for Yahoo is that no single type of data will do. "It's not about behavioral data for us. It's not just behavioral, but social data and search data and contextual data and time of day, the weather, when it rains, etc.," he says.

"Our hypothesis was that there is no single silver bullet. For some ads, some campaigns work better on certain days of the week or in some states. Search data might be the answer for one bit of the audience and behavioral for the other. But if you add them up and layer them on top of each other, you can build a better story for every ad for every person on every page for every minute." 

Which begs the question I have heard some planners ask more and more: Are we at risk of over-automating? Will this all come down to math in a black box? Is this a complexity that necessarily excludes much human intervention?

Frankel is blunt. "I think they are right to be nervous. Planning is in for a fundamental change. I think manual planning of every impression is probably outmoded. It doesn't mean that planning is dead but that it has to get smarter," he notes.

"Of course there is room for something other than the math," he adds. "But I think there is a lot more room for the math and for the people who can extract values ultimately, in ways that are going to look like a black box to somebody. But the problem is that if you believe in my theory -- that there is no one kind of data that is the perfect data, and it needs to be added up it, it needs to be geo, plus time of day, plus creative plus what site you are on -- no human being can manually organize all of those different options. It is going to take machines to figure it out."