Behavioral Targeting Distortion

Sometimes I wonder what my online profile says about me and whether there’s a place I can go to determine if online marketers have me pegged appropriately in my many roles at home and work, in my community, and probably in a few other places. There ought to be a credit bureau of sorts where you see if they got it right and correct it if it isn’t.

This really isn’t a possibility because the behavioral targeting networks responsible for the bulk of the behavioral traffic operate, for the most part, independently. And most consumers wouldn’t bother. But even if consumers did check this mythical information clearinghouse, would their own perceptions be any more accurate or valuable to marketers?

Truth is, while we all may wear multiple hats, we wear a dominant hat at any given moment. To be really effective, advertisers must read and respond to immediate past cues to be relevant to users’ hat of the moment. The best publishers develop algorithms and practices that do this pretty reliably, but they still have to filter out a lot of garbage before it pollutes their profiles. The trick is identifying what is, in fact, garbage.

We already know that seasonality is an issue that affects longer term behavioral data. Consider the holiday gift-giving season, where people are buying things online for other people and not necessarily indicating their own purchase interests. There are behavioral segments that have finite periods of relevance, such as buying a car, planning a trip, or planning a wedding. Much depends on the appropriate cookie window to narrowly focus efforts in the right direction.

I recently had a terrible site experience and reloaded a malfunctioning page 15 times, spending an inordinate amount of time on that frustrating site. I wondered what that said, if anything, about me to marketers. I’m guessing they didn’t put me in the frustrated and annoyed willing buyer segment, though it would have certainly been accurate at the time. While interests and behaviors may expire, we would not call them distorted if they reflect a real and current need. Our needs change constantly, and so must our profiles to be an accurate reflection of our behaviors.

Recently Jeff Hirsch, CEO of Revenue Science, broadly defined three common types of behavioral targeting and made a case for the Boolean logic approach that his company practices:

  • Clustering. Most of the networks that offer behavioral targeting are doing this. They tag many of their sites. When a user goes to a sports site, that person is dropped into a sports bucket. Someone visiting a travel site gets placed into a travel bucket. A person can’t be in two different buckets. Very little behavioral data goes into segment creation. When you buy the network’s travel segment, you’re getting a pool of retargeted users who visited a site one time.
  • Predictive modeling. There are a few ways to approach this, but it generally starts with a test using no targeting to collect data. A behavioral targeting provider may collect data to identify people who travel based on sites they visit. The provider then looks at other behaviors these people exhibit. This establishes a profile. They then go find look-alikes to scale the audience.
  • Rules-based behavioral targeting based on Boolean logic. This is Revenue Science’s approach to behavioral targeting. Behavioral data is collected and stored in a data warehouse. Examples of behavioral data include articles (by topic) people recently read, search words they entered, product comparisons, shopping data (no PII (define), and much more. Then a set of criteria is entered into the database to find all the people who qualify. This is now your audience.

Maybe the better question isn’t whether we marketers are seeing the full picture but whether we’re seeing an accurate picture in the snapshot of time that matters most — right now. With behavioral targeting, we’re not really trying to determine who you are as much as what your needs are right now. That’s why PII isn’t necessary or desirable in behavioral targeting. The credit bureau analogy breaks down, because credit is an accumulation of many decisions, good and bad, that we all make over time. It might make for very good reading to capture the many faces we display online through our behaviors, but the accumulation of who we are is a mystery best left to philosophers, not marketers. For marketers it is sufficient to know how we can help consumers.

Related reading