In a recent newspaper article, a group of scientists revealed their beliefs about the diet of humans 5,000 years ago. Their beliefs were based on the analysis of the contents of the stomach of the Iceman, a prehistoric man whose remains were discovered by tourists in the Italian Alps. Based both on the kinds of food he had eaten (cereals as well as venison) and the way that he was killed (by an arrowhead in his left shoulder), these scientists suggested that his death might have been the result of rivalry between different hunting gangs. It is interesting that these scientists are discussing details of an event that took place several millennia ago, based on inferences drawn from a "witness" that had been dead for centuries.

These scientists are not alone; anyone with a healthy set of data, a couple of analysis tools and some insight should be able to reconstruct a set of scenarios based on a limited set of evidentiary elements. At a crime scene, the investigators collect as much evidence as they can – fingerprints, hair and skin for DNA analysis, mud prints. The environment of the crime scene is evaluated as well. By accumulating as much evidence as possible, the clever detective will be able to make enough inferences about the criminal to form a valid profile, which can then be used to locate the evildoer.

Similarly, archaeologists working with bits and pieces of bone or the reflection of a skeletal structure imprinted in an ancient fossil draw conclusions not just about the bone infrastructure of dinosaurs, but about all aspects of these old-time animals' lives. For example, by finding embedded seed or leaf images within a fossil, a scientist can infer a lot about the diet and biology of the fossilized creature.

Abstractly, what these scientists are doing is evaluating the evidence of some attribute or behavior and then drawing inferences about the subject based on that evidence. These inferences are then used to explore some thesis about the subject, the results of which can then be evaluated for correctness.

I'll venture to say that in some situations, people are very similar to the dinosaurs mentioned earlier. In almost every action we take, we leave behind some evidence. By evaluating the kinds of actions or the sequence of actions that people take when they need to make decisions, we can make inferences about both the way that certain decisions are made and the personality profiles of the people making those decisions.

These are two powerful concepts. The first implies that if we can identify a sequence of events that lead toward a particular action, then we can predict if that action will be taken by tracking the event sequences. This is particularly valuable when it comes to customer life cycle events. For example, at what point does someone stop shopping and start buying? Or, what are the events that lead customers to close their bank accounts?

The second concept implies that we can cluster individuals into personality profiles by looking at the kinds of actions they perform. This is interesting when we have a limited amount of resources to dedicate to some business process and we want to maximize our benefit. For example, we might make the inference that individuals who purchase fast food or gourmet take-out dinners after 10:00 p.m. five nights a week are likely to be leading a lifestyle that is not consistent with having small children. Therefore, those people would not be good prospects for purchasing minivans.

There are three challenges for exploiting this kind of analysis:

  1. Identify those actions that are of interest and create a mechanism for logging and archiving that information. For example, in a clothing store, picking up a folded shirt and unfolding it might be considered a unique customer activity that should be logged.
  2. Create an analytical framework for determining sequences of events that lead to success indicators. For example, how many times did a person visit a Web site and view a product's information sheet before purchasing it?
  3. Enable the enhancement of activity information with alternate value-added data, such as demographics, psychographics, geographic information, etc.

Unfortunately, it is relatively uncommon for an organization to have its act together enough to properly exploit this kind of analysis. In the best environment, this can lead to laser-perfect targeting. In the worst case, this leads to huge, unmanageable databases; overtaxed processing systems; and a bunch of unhappy clients. The attempt to discover "success patterns" can get bogged down with the discovery of a bunch of irrelevant conclusions that don't add value. A good approach is to start with one or two known success patterns and then formulate a pilot project to validate that those patterns can be recognized within the set of actions. This acts as a "proof of concept," and a success during this pilot should encourage expansion of this kind of analysis.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access