• by Phil Leggiere , May 13, 2009

A not-so-funny thing happened on the way to the often-prophesied friction-free online advertising ecosystem, the one where efficiency would equal ever-increasing value of inventory. While improvements in targeting technology have progressed rapidly, the value of display inventory has steadily, and steeply, declined overall. Indeed, so much so that it's become commonplace on the conference circuit to hear speculation on a permanent depression in display pricing, even for ostensibly "targeted" inventory.

A big part of the problem is what might be called "optimization silos" that still exist in the online world among content optimization, behavioral targeting and creative optimization.

Content optimization has made it possible for marketers to recommend personalized content on-site. Behavioral targeting has made it possible to identify the right audiences, and creative optimization has made it possible to create dynamic ads. Yet the data streams necessary to make these innovations work seamlessly in a scalable campaign have been disconnected.

Is there a way out of this cul de sac? A number of emerging data optimization players are betting there is. One of them, Aggregate Knowledge, is attempting to smash these optimization silos through a super-computing infrastructure that is able to ingest and leverage thousands of new data points per second, reach over 70 million users, and cover over 1.6 million pages, processing 112 trillion possible combinations per second.

With this infrastructure, the optimization platform is able to analyze each ad impression in real time and decide within a split second what content or products to show inside a site widget on an advertiser site or within a dynamic Flash creative anywhere on the Internet, Pascal Bensoussan, Aggregate Knowledge's vice president of products & marketing, explains.

"Collecting and storing data is one thing," Bensoussan says. "Actually being able to put massive amounts of data into action in real time to keep up with the pulse of the market is much harder, and that's where most advertising targeting is still stuck."

In Aggregate Knowledge's platform, personalization algorithms aggregate site behavior and weigh each anonymous user's actions, with the goal of bringing together the optimal content, products, messages, and visual elements to engage the user and drive key campaign metrics like clicks and conversions.

The personalization technology is designed in particular to work with advertisers that have a lot of traffic, large product catalogs with changing prices and availability like those of retailers and CPG companies, and a complex portfolios of offers and packages like telco and travel companies. Aggregate Knowledge is also experimenting with B2B campaigns where the advertiser has defined more than 200 ways to engage with their prospective customers (e.g., watch a video, download a white paper, or subscribe to a newsletter).

As an example of the type of impact this kind of personalized creative can have, Bensoussan cites a campaign shoe manufacturer Skechers ran to retarget past visitors, from window shoppers (people who visited the site, but did not add products to their cart) to cart abandoners and purchasers with a dynamic personalized creative and product recommendations best suited to their interests and intents. AK's personalized creative generated $7.39 in sales revenue for every $1 spent on media, compared to only $1.52 in sales revenue for every dollar spent on media with a static creative.

"We measured the lift in performance between personalized creative and a static creative designed by the advertiser," Bensoussan says, "and saw the best lift in click-conversion from window shoppers (2.9x), suggesting that personalized creative is critical to guide those visitors toward a purchase. We saw the best lift in ROAS from cart abandoners (5.4x) in line with expectations that timely and relevant offers generate the highest response from consumers with high purchase intent. We also saw the best lift in click-through rate from purchasers (3.7x), indicating that a combination of brand recognition and relevance plays an important role in bringing back past purchasers."

The next optimization silo that needs to be cleared, Bensoussan believes, is providing the agency side with visibility into where, when, how, and how often targeted ads actually reach their target successfully.

"Today agencies buy billions of impressions," he says, "most of which come to them with no information regarding the user or the context. As a result, in a typical campaign, 20% at most of the impressions come from the desired audience within the frequency cap. Another 20% might be targeted to the right people but falls outside of the frequency cap. Then upwards of 25% are non-US users and 35% are off-target completely."

The premise behind Aggregate Knowledge's argument is that to the extent that agencies gain greater visibility and control over placements and campaign targeting, they'll also begin to minimize the great hidden costs of waste that plague the average campaign.

Will this alter the current economic calculus driving the devaluation of display inventory, and subtly undermining the premium proposition of much behavioral targeting or retargeting initiatives? It's obviously way too soon to tell, but this seems like a promising and important direction.