This column, the next installment in our series on ad targeting, is dedicated to all those people who wrote me last week to say: “Tom, people like to lie on their registration forms.” I know… It happens.
We dig watching WWF Monday Night Raw and Jerry Springer, but the minute we get a marketing survey in the mail, we suddenly become fans of Masterpiece Theatre and independent movies on the Sundance Channel. So why would we expect that web-based marketing questions are answered any more truthfully?
What we do and what we say we do are two different things. That’s one of the premises behind behavioral targeting, our topic du jour.
Some web-based marketing companies are asking: Why should we subject users to a lengthy registration process and ask them a bunch of questions when we know they might flagrantly lie? Why shouldn’t we simply examine their surfing habits and deliver advertising based on observable behavior? These are legitimate questions.
There are many different ways you can target ads once you’ve built a database of user behavior profiles. One method is called collaborative filtering. That is, if you notice a correlation between people who buy product A and people who buy product B when you’re examining a sample of user profiles, you might try advertising product B to future buyers of A at the point of purchase. Collaborative filtering involves deriving product recommendations from aggregate data and applying those recommendations to individuals.
Collaborative filtering is not the only way to leverage behavioral data. Suppose a network ad management system could recognize when an individual is entering a critical period in the sales cycle. Wouldn’t it be cool to target people with car ads at the moment they start checking out car prices on the web? Guess what? It can be done.
Engage is a leading company in the space, and its profiles go beyond targeting based on content visited. Engage also takes into consideration the recency, frequency and duration of visits in its profiling technology, so your campaign can tell the difference between people who are dropping by auto-related sites to download a GIF of the Porsche Boxster, and those who are seriously comparison-shopping for cars.
One thing I like about behavioral targeting is its flexibility. You can get as complicated as you want with it, or you can keep it simple. You can launch a complex campaign targeting people in different stages of the car purchasing cycle, as in the example above, or you can target CD buyers with ads for CD carrying cases.
“But Tom,” you might ask, “how does this fit with your theory on contextually relevant advertising?” Well, it does and it doesn’t. Personally, I would bet my last ten bucks that a company advertising CD carrying cases will have a lot more luck targeting frequent CD buyers on a music-related site rather than on, say, a gardening site.
I definitely think that context has a lot to do with advertising success. At the same time, I think a CD carrying case advertised to frequent CD buyers on a gardening site is better than the same ad running untargeted on the gardening site. I’d kinda like to see an advertising effectiveness study that compares efficiencies for behavioral targeting, contextual targeting, and untargeted ads. If anyone could find it in their heart to write me a big fat check to conduct this study, I would be happy to share the results.
Some of you might be wondering how behavioral targeting fits into the online privacy picture. Engage’s profiles are anonymous, according to the company’s web site, and don’t contain “personally identifiable data, such as name, address and email address.” Fine, but remember that almost any database can be abused.
Conceivably, a web site could contract with a company like Engage to target specific profiles, drop its own cookie on users, and correlate the behavioral data to personal data it might already collect. But that would be a flagrant privacy violation. Used on its own as intended, Engage’s profiling technology appears to pass the white glove test.
Next week we wrap up the series. See you then.
