Features

Gregory Piatetsky: What is the role of data mining and analytics in EDM? How is EDM different from Data mining?

James Taylor: Data mining and analytics are core to EDM - they are how the data you have is turned into better decisions, how the system learns from its data. It is true that the basic necessity for EDM is the automation and improvement of operational decisions using a declarative, rules-based approach. However, almost all users of EDM rapidly discover that their data is one of the best sources for these rules and start applying data mining to find them - customer segmentation rules or risk profiling rules, for instance. Predictive analytic techniques are also common in EDM-based systems, with scoring of risk, fraud, opportunity, churn, propensity etc. These scores are enacted by more rules so that the prediction becomes a useful business action. The analytic approach of adaptive control, or champion/challenger, is also key to EDM and is applied both to rules-only processing and to combinations of rules and analytics.

The difference between EDM and data mining is really twofold. Firstly, EDM is ruthlessly focused on operational or transactional decision making: How should I treat this customer right now, what retention offer makes the most sense for this customer, is this transaction fraudulent and so on. Data mining can be applied in these circumstances but can also be applied to broader, more one-off kinds of decisions. Secondly EDM is focused on taking action, not increasing understanding. Data mining's ability to help you understand the past and predict the future can be used to direct and improve the actions your systems take but it is sometimes used only to increase understanding. There is nothing wrong with this, it is just not EDM. As we sometimes say, not "actionable insight" so much as "insightful action".

here is part 2 of the interview with James Taylor.