Peter Fader, professor of marketing at University of Pennsylvania's Wharton School tells CIO Insight that he hates to see companies waste time and money collecting terabytes of customer data in attempts to draw conclusions and predictions that simply can't be made.
He's got an alternative, however: Complement data mining with probability models. He notes that data mining can be good for certain time-sensitive analyses, but he does not believe they are as useful for making specific forecasts about what particular customers are likely to do in the future. For that, he says, different tools are necessary. "People keep thinking that if we collect more data, if we just understand more about customers, we can resolve all the uncertainty. It will never, ever work that way." He goes on to discuss probability models as an alternative to data mining. "Probability models are a class of models that people used back in the old days when data weren't abundantly available. These modeling procedures are based on a few premises: People do things in a random manner; the randomness can be characterized by simple probability distributions; and the propensities for people to do things vary-over time, across people, across circumstance.
For more:
- read this probability models article
