• by Kate Kaye , December 9, 2005

Like most of his marketing industry brethren, Bill Harvey has long considered contextual advertising to be the holy grail of effective ad targeting when it comes to branding, click-through rates and ROI. But a number of recent case studies measuring the outcomes of behaviorally targeted campaigns against contextually targeted campaigns for advertisers including BMW and Snapple has Harvey, the president of research and consulting firm Next Century Media, reconsidering this commonly held belief. Harvey's company has teamed up with behavioral targeting outfit Tacoda Systems to conduct an extensive series of studies analyzing this unexpected phenomenon. Behavioral Insider spoke with Harvey about his initial findings and future research plans.

Behavioral Insider: In a recent presentation you gave, you cited many examples of behavioral targeting (BT) outperforming contextual (CT), run-of-network (RON) and demographic targeting. It seems like the running theme in that presentation was that advertisers will find such results surprising. Why would they be surprised?

Harvey: Well, it's not surprising that BT would beat RON. But there are two media variables of effectiveness. They're the target and the unique effect of the environment. Between BT and CT, the target is the same, so the only variation is the editorial environment. In that situation, you have the choice of reaching the target (let's say, people who are interested in buying a new car) either at a site that's all about cars, or reaching them in some context that has nothing to do with cars. You'd think that it ought to be more effective in the context of people who are there for the purpose of making a new-car-buying decision.

BI: And it seems that's been the consensus throughout the history of advertising.

Harvey: Absolutely. 'Editorial environment' was what we called it in the magazine field. I'm a great believer in it; I've done 28 studies to show the value of it on the Internet in terms of Internet sponsorships. So, suddenly to find that no-context has more effect than context, that's the surprising part. And incidentally, it doesn't always happen this way. It's only that 25-50 percent of the time [we're finding BT is more effective than CT].

BI: What are the reasons that you hypothesize are behind the better effectiveness of behavioral in these cases?

Harvey: We've got two hypotheses so far. One we call 'clamor.' Clamor is, for example, [when] you go to a car site and there are all these car ads competing for your attention. Also, there's strong car editorial. So, it's possible that some people coming to the site are drawn to the editorial, or maybe their eyes are even avoiding all of those competing car ads.

BI: Do you consider clutter to be the same thing?

Harvey: Yes, clutter is the generic term for it, but as it affects this specific question of BT vs. CT... we're calling it clamor.

And then the other hypothesis is what we call 'surprise.' Let's say you've gone to fashion sites, so you're behaviorally targeted as interested in fashion, and you go to a baby-care site and you see a fashion ad. You're not expecting that, so you're surprised.

....What we've learned... is [that] our brains are hard-wired to a survival program to form an expectation, a model of what we're going to find in any given environment. To the extent that what is actually in the environment meets the expectation, we don't notice it. But then when something isn't expected, the eyes are drawn to it, and in the brain there's a potential [cerebral cortex] wave called the P300 wave.

....For example, we're doing a test right now with Panasonic... an InsightExpress 'traditional' branding study, and we're also doing an eye-tracking study with the same Panasonic ads (there are five different units). So, the eye-tracking study is to see whether eyes are avoiding the Panasonic ads in CT or not. We're planning to do a brainwave study to see if in behavioral targeting there's a P300 wave indicating surprise, and if the eyes are drawn to the surprising element.

BI: So there's a real human factor to these studies as opposed to just tracking online interaction.

Harvey: Well, we want to know the 'why.' We're getting the facts about the interaction, we're'getting the facts about the purchase intent, all that's great.... But the idea of BT was never to replace contextual targeting. It was always to extend the reach and frequency of contextual targeting; often with contextual targeting, the inventory is sold out.

BI: They've kind of worked hand-in-hand at this point.

Harvey: And probably always will. It's our intent to make both CT and BT better and to learn how to mix them in different contexts. There's no competition.

BI: Do you see pockets of verticals or environments that might be more conducive to behavioral working better than contextual?

Harvey: Over the next year we want to collect as many case studies as we can; we've asked the industry for their case studies. We're going to look at them by the dimensions you're talking about to see under what conditions we get BT beating CT in branding and click-through and ROI and what's the optimal mix. One of our aims is also to see whether CT can be improved. If clamor is deleterious, under what conditions can we fix it? Can we make it economically viable for both buyer and seller for there to be more solo sponsorships, or restructure pages or ad units so the clamor effect is minimized?

BI: Over what period of time will the study be conducted?

Harvey: It'll be over at least a year. There are really many studies involved.... The Internet has now gone through the original enthusiasm and then the bubble, and now it's in a mature phase of rapid growth. Unlike the traditional media psychological approach between buyer and seller which was always adversarial, now, to some extent, Internet sellers are more interested in helping buyers achieve effectiveness and approach things more as a partnership, not as a fight. It's not so much negotiating, it's more about success together. If we can contribute to that by doing the best research, then that's a good thing.