The core preoccupation of online targeting thus far has been to scout out from among the millions of active Web searchers, shoppers and browsers the most likely prospects to pursue. The next step in the pursuit of maximizing ad relevancy, Toby Gabriner, CEO of x+1 argues below, is to explore the other half of the targeting equation--identifying and screening out the consumers who don't want to see your ad.
Behavioral Insider: What are the main limitations of conventional behavioral targeting?
Gabriner: Behavioral targeting gives you one attribute. It tells you someone is in market, but it doesn't tell you who they are and what they are in market for, what they are willing to pay for, and what kinds of messages they're likely to respond to. You don't know anything further about that individual, like their age, gender, or geography.
Behavioral targeting gets and deserves a good deal of credit for enabling publishers to identify and segment groups of users based on past traffic patterns. Once advertisers identify the audience they are looking to reach, publishers can sell this inventory accurately. But, the problem with this model is it is not as fluid and flexible as the Web appears to be. The experiences and the profiles of the audience change quickly and are reactive to the environment around them, so the past behavior of an audience, though relatively accurate, has its limitations. We do need to look at behavioral and contextual variables. But we need to work as hard in separating out non-prospects as at finding prospects.
BI: How does what X+1 is doing with Audience Screening work?
Gabriner: Audience Screening allows the advertiser to identify the audience represented from an impression on a network or a portal and determine if that audience member is more or less likely to act in response to an advertisement than the general audience. If the audience member is regarded as highly desirable, then the ads are exposed. If the audience member is not deemed highly desirable, then they are not exposed to the ad.
Think of it as collaborative filtering, but on a wider scale. If you're a site owner or an advertiser, we have the ability, when someone comes into a Web site, to locate the new visitor with different data points within a relevant cluster.
BI: Who have the early adopters of this mode of targeting been?
Gabriner: This type of targeting is getting its first take-up from direct response advertisers, because screening has its most immediate and immediately apparent benefits there. If you're selling dog food, to take the basic example, you want to screen out all [those who don't own dogs.]
Financial services, teleco, auto, travel, e-tailing and hospitality are at this stage the first adopters, but we see this methodology fairly seamlessly extending into more branding environments.
BI: What types of ad campaigns and goals is it most suited towards?
Gabriner: Well, let's take telecommunications. The number of products and services that need to be marketed in order to be competitive has increased dramatically. So many providers now have a full suite of voice, IP and cable services, with all sorts of segmented product or service plans to meet the specific needs of consumers.
But the problem for contemporary marketers is, customers do not always fall into easily defined segments or profiles easily identified solely by their behavioral profile.
What the technology does is take profiles of the different audience segments advertisers want to reach for specific ads. Say it's a male 25 to 34 with a propensity to spend money on certain online site categories. When an impression is served by one of the ad networks, the screener reviews it in real time and analyzes the behavioral profile of the consumer who could receive it--and then checks that, in addition, against demographic, geographic and other criteria to make sure the match between message and the prospect profile is strong. If it is, the ad will be served, but if the match is weak the ad will be screened out.
BI: How does it make these kinds of predictive decisions?
Gabriner: The Audience Screen model actually identifies this information in real time and can be updated faster and with more detailed accuracy. Audience Screening takes into account audience profile IP and header data, preferably in conjunction with industry reliable sources such as Claritas or Simmons, and merges this with data referring to the page where the ad is shown, the category of the site and more recent events (i.e. news, etc.).
BI: What are the main challenges and opportunities for BT in this space going forward?
Gabriner: It's working if it allows you to target the long tail, tier-two sites as effectively as the short tail.
Right now it's limited to display only, but on the horizon we see no reason not to extend that to optimize keywords, mobile, e-mail and anything IP-based.
The payoff for audience screening is that publishers can help media planners be more efficient. If you can eliminate audiences members who will in all likelihood NOT respond to the advertiser's message, why not? If you can adjust the client's messaging on the fly to assure a higher response rate, why not? It will improve the client's ROI and your standing in their eyes.