• by Phil Leggiere , December 19, 2007

The operating, usually unconscious, assumption of most behavioral targeting programs is that "you are what you click." Attempting to understand consumer's behavior strictly by their history, however, as Branton Kenton-Dau, director of VortexDNA, explains below, is not only poor psychology but unproductive marketing.

Behavioral Insider: VortexDNA eschews the "historical" approach to behavior most of us associate with behavioral targeting. How exactly does your methodology differ?

Kenton-Dau: There are several challenges to optimizing behavioral data using traditional methods. One of the biggest is that it has to be historical. If you've searched for beetles, I still wouldn't know whether you are more likely to prefer deep-sea diving to surfing. So my ability to optimize is limited to the size of the historical picture of your behavior over time. And that picture raises privacy questions: how you treat the data, how you protect it, who you share it with, all those things.

The VortexDNA platform takes an entirely different approach. Because it's based on core beliefs rather than click stream, it is predictive without being historical, with no privacy concerns. The only data that is stored is an aggregated 7-digit number, which is predictive in a far more profound way, with no limitations of genre.

This is not to say that the traditional models are bad or wrong. There is, however, a definite gap that is filled by a non-historical model.

BI: What was the motivation for developing your technologies?

Kenton-Dau: The original spark for this work came from the book ‘Built to Last,' by Jim Collins and Jerry Porras. In that book, they look at a group of companies that outperform the stock market, and make the claim that strong alignment around a shared set of values is one of the distinguishing features they shared. We [Branton Kenton-Dau and Raf Manji] decided to validate this concept against 750 companies in the S&P 500 and the NASDAQ, and found it to be true: strong-culture firms have a yearly return 7% higher than weak-culture firms on average, which doubles their returns to shareholders every 8 years or so.

Our work studying these concepts led us to the understanding that human intention, which guides things like purpose and values and the shared culture of these companies, is governed by the mathematics of complex systems, and it has predictive characteristics. This means that we can calculate someone's purpose and values mathematically, and compare it to anyone or anything else to get a measure of alignment.

Our model completely reinforces and complements existing demographic, contextual and behavioral targeting -- these models work by studying the outward ‘symptoms' of someone's core intention; our model works by adding information about the inner cause.

BI: I'm wondering if you could sketch out a little further how exactly the system goes about inferring core values and intentions.  I'd love to see if the concept (which is quite unfamiliar, I think, to many media buyers) could be fleshed out a bit more for the uninitiated.

Kenton-Dau: There are two ways that we can generate a values 'genome' for an individual user. One is through simple surveys, as many as the user wishes to take. The questions are designed to calculate the user's coherence (who you are) and optimization (how you act). The more surveys a user completes, the richer the profile will become.

Realistically, though, we are well aware of the difficulties in getting a high percentage of a given population to complete a survey. The second way that we generate genomes is through inference. Each site you visit has its own genome, and we can aggregate those to infer your personal genome.

It's really important to note here that we're not tracking click-stream; we're updating numbers. Let's say your genome is 1 (I'm using single-digit numbers for simplicity -- the actual genome is seven digits). You visit a site with genome 5. Your genome becomes 3. We have a more accurate picture of who you are, without having to store any information about which sites you visit.

BI: What, in your experience, are the most misunderstood things about behavioral targeting by publishers and advertisers just beginning to develop behavioral targeting features?

Kenton-Dau: The belief that there is only one way - to track user history and other user characteristics.

BI: Could you elaborate a little on how your approach might be used in tandem with more traditional behavioural, demographic, contextual approaches?

Kenton-Dau: Because VortexDNA operates as a predictive filter, it can be combined with any other approach to further enhance and refine results at any stage of the calculation. It's basically one more piece of data: if you live on this street, have this many kids, make this much money, and have this VortexDNA profile, you're more likely to enjoy this product. The big difference, of course, is that the VortexDNA piece of data is the only piece that is truly personalized -- just because people live on the same street doesn't mean they share the same values.

The same holds true for behavioral data, in particular because we can infer a genome for users based on their behavior. Because it's non-historical, though, it can be a lot more powerful. Let's take Amazon as an example. Amazon says, 'People who bought this book also bought this book.' They're relying on people actually having bought both books. If they overlaid VortexDNA onto their recommendation system, they could also say, 'If you like this book, you might also enjoy this one,' even if nobody has ever purchased both books together.

BI: Can you give some examples of your behavioral platform is being successfully deployed?

Kenton-Dau: We are bound by NDA, but we have two direct marketing companies in paid trials using the technology - a $3 billion and a $300 million company.  We also have two paid trials underway in the interactive TV market.

BI: What are the most important challenges to effectively deploying and growing the platform you see on the horizon into 2008?

Kenton-Dau: Serving market demand for a proven technology that is easy to use and applicable across all enterprises seeking better behavioral targeting.