SAS STRIKES NEW DATA MINING LODE
By Mark Hammond

SAS Institute Inc. is driving its data mining efforts down a new road: vertical industries.

The Cary, N.C., developer will announce this week an upgrade to its Enterprise Miner tool, Version 3.0, that offers templates for cross-selling in financial services and for actuarial rate-making for the insurance industry.

In addition to the cross-selling and actuarial rate-making features, SAS officials said the company is developing templates specific to financial services, retail, telecommunications, utilities, manufacturing and electronic commerce. However, they declined to give ship dates.

Data mining tools use algorithms for exploratory data discovery, pinpointing patterns in massive data volumes that are otherwise difficult to detect.

The Enterprise Miner upgrade comes at a time when demand for such tools is on the rise. New features, such as fraud detection, customer relationship management and e-commerce traffic, are expected to drive that growth. International Data Corp., of Framingham, Mass., pegged the data mining tools market at $259 million last year, growing to $1.78 billion by 2003.

"The next generation of data mining is going to be more focused on specific business applications," said Rich Rovner, analytic solutions program manager at SAS. "We are moving very quickly in that direction."

Beyond the application-specific templates, Enterprise Miner 3.0 will feature a streamlined interface and a new component, called Target Profiler, that uses predictive modeling to forecast profits.

Version 3.0, which will ship this week, also has stronger visualization capabilities, enabling users to see the results of data mining projects graphically, Rovner said.

Another new feature will enable multiple users to simultaneously work on the same data model and allow multiple models to be combined, speeding the execution of large projects.

"That's a huge feature," said Enterprise Miner user Erin McCarthy, manager of statistical services at Vermont Country Store, a catalog retailer in Manchester Center, Vt. "Sometimes if I do a really deep decision tree with lots of branches, I'm writing down 20 or 30 different variables, and that's time-consuming."

While Enterprise Miner is not yet Web-enabled, Version 3.0 introduces the capability to output results in HTML and to render results in easy-to-comprehend text.

"Before, you had to look at the numbers and figure it out," McCarthy said. "Now it's going to output the results in English, so you can just print it out and bring it to management."

Vermont Country Store uses Enterprise Miner to be more precise in targeting customers with its 18 annual mailings that total 40 million catalogs.

The return on investment is substantial, McCarthy said. "We have the luxury of really focusing on improving our dollars-per-book, measuring the average order for each catalog we mail out," she said.

McCarthy expects that ease-of-use enhancements in Enterprise Miner will help to broaden the use of data mining.

"Anything that makes it easier to train new people coming in, I think that's key," she said. "People can now come out of college and use these products and get good results with only a month of training, and going forward, that's going to be critical."

SAS' move comes as database heavyweights delve deeper into data mining. In recent weeks, Oracle Corp. bought data mining pioneer Thinking Machines Corp., and Microsoft Corp. launched an effort to build a standard mining API. In addition, IBM is weaving data mining into its DB2 database.

Pricing for Enterprise Miner 3.0 starts at $80,000 for five clients.