Plumbing the depths of user history information, connecting a myriad of dots in search of an ever-more comprehensive profile behavioral targeting, may be becoming too clever for its own good, Meir
Zohar, CEO of behavioral data exchange eXelate explains below. Rather than reveling in hyper-dimensional geometric complexity, the smarter, as well as more privacy-friendly marketing move, he
suggests, is to stick with the simple basic premise that the shortest distance between two points is a straight line. Behavioral Insider: Your platform has been advanced as an
alternative to more intrusive forms of behavioral profiling. How do delayed ads work?
When we designed the architecture for what we call event-based targeting, we
thought privacy would be the issue and that the whole now-dominant paradigm of user-profile based targeting would become more and more problematic. Basically delayed ads begin to be generated based on
particular verticals, vertical sites or vertical-specific search queries. At that point a partner ad networks bids for permission from the publisher and then sets up a cookie, which essentially
identifies a particular user as a target for an ad relevant to their interests. Over the next few days, when the network has available ad space on one of its sites in that vertical, the delayed ad
cookie is identified and activated, and the delayed ad is served. But, quite importantly, there is no user profile created. The cookie is not based on cross-site information and integration. It's
strictly limited to a single event and a single site visit or search.BI: So there's no user tracking, or connecting the dots, as we'd say?
With event-based targeting, you don't follow and track everywhere users go online. You don't aggregate behavioral patterns to create a profile. All you do is simply use a cookie to associate a
user with one particular event, a visit to a vertical site, for instance, and then to use that data to serve a subsequent ad when that user visits a similar content vertical site. A simple example is
that someone visits an international travel site related to Spain. When that user subsequently goes to a travel vertical on the ad network which bid for the delayed ad cookie, [he or she] can be
served an ad for Spanish travel.BI: What verticals are you currently working with?
We have a number of vertical channels currently, including
travel, small business, technology, Hispanic -- and we will be steadily increasing the range of channels.BI: How does a network participating in the platform go about doing a
Say we have an ad network on the exchange. They essentially purchase directly from publishers the right to set up a cookie on a user visiting their page. The cookies
record their interest based on the vertical they're visiting, but that's all they do. Then say about five days later a site on their network within that identical vertical has available
inventory, they can serve an activated delayed ad to a customer with demonstrated interest in that vertical. When you have vertical data bid for on an exchange basis, it maximizes the value both of
each data piece and where it is served.BI: What about the value proposition for advertisers and publishers?
We believe going forward that keeping a loyal
customer base will become ever more crucial in monetizing sites, especially social networking sites where member base is the basis of valuation. For publishers, of course, it's a huge win-win,
extending the possibilities for monetization beyond selling ad inventory on-site. For advertisers, it's another layer of visibility in having ads served over an ad network. For ad networks it
radically expands the possibilities of attaining targeting efficiencies, beyond those feasible within the bounds of the network.BI: Beyond avoiding privacy issues, which is
obviously a great plus in and of itself, what kinds of increased efficiencies can you point to for event-based targeting?
By sticking with live highly actionable data,
delayed cookies avoid another pitfall of behavioral targeting -- which is that it aggregates so much information that by the time it's used it's become stale or outdated. There's often a
very short shelf life for many aspects of behavioral data and a sort of law of diminishing returns can set in. By focusing in a laser-like manner on discrete events, we believe we can maximize