Thursday, November 01, 2007

Predictive Analytics and Content

Content management incorporates many prinicples into a paradigm. Any implementation of these principles may be more or less cohesive depending on how well the business processes are understood and how well the underlying functionalities are customized to support and extend these business processes.

A relatively recent benefit of content management is predictive analysis. Predictive analysis is an example of impelementing business intelligence into an information lifecycle. This particular BI hinges on correlating seeming disconnected pieces of data using heuristics to produce a probability matrix of knowledge entities from a vast respository where those entities are most likely to fit some criteria. An example is that Infinity Property and Casualty, an automobile insurance company, recently selected SPSS, Inc's Predictive Claims (tm) software to streamline their claims process (link). In this instance the company suggests that predictive analysis will allow them to better assign claims to adjusters best-suited to handle each particular claim with a side benefit of helping their company detect fraud more efficiently.

I believe that most people probably don't tend to think of their insurers as penny-pinching fat cats who eschew handing out a dime but if you are one of the minority that fits into that description here's something that might help. Handling claims costs money and while there are some instances of egregious behavior through the insurance universe, there are also myriads of stories of claims reps in any insurance industry that make heartfelt, sincere efforts to help their claimants throughout the claims process and get the insured customers' funds out to the customers in a timely manner. Unfortunately, just like there may sometimes be a person or process or unspoken policy that hampers even the best-intentioned claims rep, sometimes there are unscrupulous individuals who might fake an injury or have an estimate given by a "friend" to defraud the insurance provider. Predictive analytics is a hedge for the insurance industry against just such fraud. This helps you and me (assuming you're not one of the people perpetrating fraud). Believe it or not, in the same manner that an actuarial can often accurately predict who is most likely to have a wreck or get injured (which is how insurance companies determine what your rates are for your insurance), predictive analytics can very accurately narrow the pool of potential fraud perpetrators so that the limited investigative teams of insurance adjusters can spend their limited time checking out the other guy who claimed his late-model Lamborghini got a scratch at the mini-mart so that our car insurance carrier doesn't waste too much time having an adjuster come out to your house to see how your truck got totaled by the dude rushing to get to work last week.

Sometimes it can be annoying to think that anyone has this much information and the ability to use it to predict things but its important to keep two things in mind about predictive analytics:
1. The only cases I've heard of are where predictive analytics is used to create a matrix of people for someone to check into; no one is ever convicted based on predictive analytics. These tools are used to better assign resources.
2. Just because the probability that you will do X is higher than the probability that I'll do X does not translate into the fact that you'll do X before I will. People have a capacity for creativity and while how someone operates based on their upbringing, environment and heredity is for another discussion, it is eminently clear that content systems do not contain all possible permutations of events that might contribute to a decision that you make so a million computers running predictive analytics could analyze volumes of data about you and they could all reach the same conclusion that you're going to be in a car wreck next Tuesday and yet, through the sheer carpricious nature of chance you might not be in a wreck. Just because something is predicted, even based on these impressive predictive tools, does not mean that it has to happen.

Hope this gives you some insight into predictive analytics and content management from the 20,000 foot view.

For a more educated view of predictive analysis, see the 24-page White Paper produced by that AICPCU/IIU - link.

No comments: