But there are two problems: It's not pervasive yet, so reach is an issue. And segment characteristics from site to site don't match, causing inconsistencies. The lack of standards, which causes the inconsistencies, is of particular concern to Nick Pahade, VP and managing director of Beyond Interactive. "...Everyone's methodology is different, as is how they're defining behavior," he says, "particularly with the IP address. How can you use the IP address as a defining factor when one address could be shared among 30 people? It could be a household. It could be a school."
These are short-term problems, however. First, major growth is imminent. Not only are publishers being pushed by agency demand, but the revenue opportunities are so obvious that they're pushing themselves. Second, suppliers are coming up with unique solutions that address both the lack of reach and the inconsistency issue.
What BT Does
Imagine this scenario. Your client wants to reach travel enthusiasts, but inventory in travel sections is seldom available. So you buy what you can on as many sites as possible. Eventually, you reach your intended number of impressions, but you end up with a very fragmented plan. Meanwhile, even though you are advertising in the travel section, you have no proof that the people you are reaching are actually travel enthusiasts.
Using behavioral targeting, a site categorizes visitors into groups based on non-identifiable parameters, such as clickstream behavior over time and IP information. For instance, a travel enthusiast might be someone who visited the travel section four or more times in the last 30 days and also clicked on a link for more information about a trip. Now, a person who matches the description of a travel enthusiast visits the travel section, but your ad doesn't appear because the section's inventory was sold out. A few minutes later, that person moves on to the finance section, where your ad is delivered. Are you reaching your intended audience? Absolutely. Even though he is no longer looking at travel information, he still meets the criteria of a travel enthusiast. You just found more Web visitors who match your needs, and the publisher sold additional inventory. It's a win-win for both sides.
How it Works
"There are three types of data we capture from a publisher," explains Omar Tawakol, director of marketing at Revenue Science. "What people say - registration information, searches, etc.; what they do - their clickstream data, where they're going and what they're consuming; and how they do it - IP attributes, geography, are they coming from work or home," he says.
"The first two sets are gathered via a cookie and a pixel at a publisher's site. A pixel is an invisible 1x1 image that creates a log of activity in a seamless, autonomous, and real-time way," says Tawakol. In addition to tracking the current visit, cookies and pixels also remember behavior from previous sessions.
There are three components needed to make behavioral targeting work: 1) analytics; 2) the behavioral targeting piece itself; and 3) an ad server.
At the top of the chain, an analytics tool studies the publisher's Web-visitor behavior and determines what classifies a person as a travel enthusiast, or an individual investor, or someone who is ready to make an automobile purchase. At the bottom of the chain is the ad server, which delivers the ads. In between the two is the behavioral targeting tool, which matches current visitors to predefined segments and then tells the ad server who qualifies for which ad.
Technology developers Revenue Science and Tacoda Systems have the bulk of the market share, while last November, Accipiter entered the market. Both market leaders have a strong list of online publisher clients. Revenue Science works with The Wall Street Journal, Financial Times, and Reuters, and recently signed ESPN. Tacoda Systems works with The New York Times,The Chicago Tribune, and USA Today, among others.
Now available: Consistency, compatibility, and scale
In recent months, four new initiatives have launched in the behavioral targeting arena that address advertisers' and agencies' concerns, including advancing scale, and eliminating incompatibilities and inconsistencies.
Advancing scale. Ad networks from Aquantive and Tacoda Systems, empowered with behavioral targeting technology, allow advertisers to increase their reach by making a large-scale buy, as well as improve their targeting ability by having more sites to choose from and being able to pick only those sites that best meet their needs. Aquantive, the parent company of online media buyer Avenue A, launched a behaviorally targeted ad network called Drive Performance Media, engined by software developed by Atlas DMT, another Aquantive division. Through Drive, advertisers request the audience they want to reach and bid a cost-per-action price on what they are willing to pay for that audience. Advertisers get a list of the top 250 publisher sites as ranked by Media Metrix, and can opt out of the ones they don't want their ads to appear in, but can't choose the sites they want, nor do they get a list of the sites they ran on afterward.
With this auction-type model, advertisers lose an element of control, but there are benefits. "We have packaged-goods clients saying, 'I couldn't before use the online channel effectively because I'm selling a product only 5 percent of the world is interested in,'" says Scott Howe, Drive general manager. "We also work with folks who say, 'I want to sell an airline ticket, but I only have West Coast routes, so I only want people who live in the western part of the U.S.'" One client that ran geo-targeted ads generated lift of a couple hundred percent over previous campaigns. "When we layered in inferred gender, it performed five to six times more effectively," Howe says.
With Tacoda's Audience Management System launched in late April, "The user has a dashboard, and from that dashboard, he can pick audience segments to serve an ad to, the kind of ads he wants them to see, set a frequency cap, and direct the ad anywhere he wants to. He may say, 'Serve them only if they're on this page'," says CEO David Morgan. Advertisers also have the option of taking other third-party or Aquantive data into account, such as demographics.
Similar to Aquantive's Drive, advertisers purchase an audience based on a pay-per-click bid model, and can't choose specific sites. The pricing gets its origins from the successes of search, which is a model advertisers may prefer. "I believe charging per action on premium targeting probably makes more fiscal and mutual sense, due to advertisers paying only for quality responders and ensuring our return business," says OMD's Finnegan.
Tacoda advertisers can reach a current base of 20 publishing firms, including Scripps Network, Belo Interactive, and USA Today. While most search prices are in the 35 to 50 cent cost-per-click range, according to Morgan, "We expect this product to price in the dollar-plus range," he says.
Technological improvements. In our earlier example, we suggested that a publisher might consider someone a travel enthusiast if he visits the travel section four times within 30 days. But another publisher might say it's someone who visits six times, or may bring other, different parameters into play. In fact, if you purchased ads against travel enthusiasts at a half dozen sites, there likely would be differences in all their definitions. Revenue Science is attempting to make those differences seamless by letting its Audience Search technology make the decisions. Audience Search adds natural language words to the traditional RFP information agencies present to publishers. So, "Instead of needing to understand the raw data, you could say, I want business travelers that fly frequently to Asia," says Tawakol, "and the software determines the characteristics of that kind of person and seeks them out." Or, adds CEO Bill Gossman, "say I want to find college-educated soccer moms in the Northeast. You're still going to ask for them that way, but now you can add psychographics, such as, 'This person is socially active and interested in family matters.'"
The technology stores and sifts through months of data, but publishers specify how far back they want profiles to be built on. Profiles are continually updated in real time. The tool also includes a behavioral relevance score - "A slider bar, with maximum reach on the left and maximum relevance on the right," Tawakol explains. "With it, you can figure out the optimal point for your advertiser." In addition to creating packages around sold-out sections, publishers can also prepackage a target audience for a group for which there is no existing editorial focus, such as the pharmaceuticals industry, or can customize based on a specific client's campaign goals.
On March 18, 24/7 Real Media introduced a behavioral targeting tool as part of its Insight suite, which eliminates the incompatibility issue caused by trying to match up the three different pieces of software. Called Insight ACT, the technology sits in between the firm's analytics tool, Insight XE, and its ad server tool, Insight OAS, to form a complete three-tiered package that speaks one language.
Publishers that are part of 24/7's Web Alliance ad network will be asked to opt in to whether they want to offer behaviorally targeted ads to their advertisers. Publishers who acquire the technology outright from 24/7 have a choice of either purchasing the software and hosting it on their own networks, or leasing via 24/7's ASP offering. Ad-network publishers that have opted in to to date include Investors.com, Lycos and Women's Forum.