So, I'll attempt to define these two terms using words B2B marketers can understand. Data Mining is using technology to look across large transactional data sets--almost always organized around customers or leads--and build executional models. Executional models are algorithms that are re-applied back on databases as scores, segments, estimates of opportunity, etc. This is key--executional models tend to be much less abstract than a model that would be developed by, say, an academic.
Data Mining is different from simple Statistics in a couple of respects. First, it tends to rely more heavily on certain types of models, such as logistic regression or CHAID--models that tend to be focused on predicting an event or classifying cases into groups. Most statistics packages (like SPSS or SAS base) have some but not all of these algorithms. Conversely, most statistics packages have algorithms that data mining platforms do not have.
Second, Data Mining programs have very rich data manipulation and transformation tools embedded. You can do everything Clementine does using SAS code, but it would take you a long time and a lot of Data step programming. Most importantly, you can integrate directly with an underlying relational database. You don't have to pull out a flat file to do your work--you can query and model in one integrated step. This is key.
In short, Data Mining allows marketers to quickly classify and analyze large volumes of customers and prospects in a relational database using more intuitive user interfaces than a programming window.
Predictive Modeling is related to data mining, and the buzz factor on this is at an all-time high. It is the collective set of tools that allow companies to predict what will happen based on historical and current period data. This is nothing new; it is essentially econometrics and statistics. What is different is that data is now making it possible on a large scale.
So how do these technologies / tools impact B2B marketers? They are being used today mainly inside of B2C organizations. Banks use them to detect fraud, direct marketers use them to score lists, and DHS uses them to find terrorists. However, they are also incredibly powerful tools for B2B marketers. Here are a few potential applications:
-- Forecasting lead quality based on previous lead behavior and prospect information
-- Finding high-potential partners based on past partner behavior and cohort analysis
-- Understanding next logical products for customers to improve sales force effectiveness
-- Modeling pipeline dynamics and using this model to build better forecasts....
The list goes on and on. There's probably another post that needs to be written about all the tools out there and what they're good for... I'll save that for another time. I'm also going to write another post on the Analytics IT Landscape Normative View and put out a straw man model for how marketing orgs should think about technology... a brainstorm I had during one of the sessions this week.