A better banner

So how can you take the “customized banner” concept and roll it forward to the point where it’s so accurate the prospect in question is nearly compelled to buy on the spot? There are technologies out there that are particularly effective at anticipating consumer needs, although not many of them have been widely deployed. One methodology that works surprisingly well in predicting consumer response is psychometric profiling. This is essentially the merging of statistics and psychology; it’s hard to predict what one person will do in a given situation, it is much easier to predict what 100,000 people will do. Behavioral parameters tend to lock in once you’ve reached a statistically significant sample (and it’s not much, statistical validity can be as little as 3%). A simple example of this would be polling data on a news site; go to CNN.com in the morning, there’s usually a polling question there—e.g. “Which of the following statements do you agree with—Bush is a jerk, or Bush is way cool.” The percentage response to each option when 100 people have responded early in the morning is almost exactly the same later in the day when 5000 people have responded. Crowd behavior is easy to predict from a statistical modeling point of view. So if you take the same statistical approach and apply it towards value-oriented statements and ask for a response from a large group of people, the same types of patterns begin to emerge from the data.

An example of this would be using a navigational tree structure to run a simple set of questions across a statistically significant sample. What I mean by navigational tree is that the questions are multiple choice, the answer to the first question determines the second question asked, the answer to the second determines the third, and so on. Since each question has e.g. five answers, the number of possible “answer paths” an individual would follow begins to increase by a factor of five with each additional question. If there are a total of eight questions, with five responses each, but each answer triggers a unique follow on question, you end up with a total pool of 390,625 possible answer combinations that a person could move through as they answer each of the eight questions. If you then track the question path through the tree structure, what you end up with in the end is a psychometric cluster populated by people who have the same value system (I would also point out the questions are incredibly important—when we did this before we hired very expensive psychometricians to develop the questionnaires). The people in the clusters may have wildly divergent demographic profiles, but they all think the same(that is, they have the same values). This type of information in the hands of an advertiser takes ad copy out of the realm of who, what, when, and where, and into the realm of WHY, which is a much more interesting way to approach a potential prospect. More on this later.