INTEGRATED DW PRODUCTS INCREASING IN POWER & POPULARITY

PC Week has noted that IT executives are increasingly utilizing emerging software suites that combine data mining tools with a data warehouse and vertical applications to more effectively predict customer behavior.

Vendors such as Informix Software Inc., SAS Institute Inc. and IBM are enhancing their data warehouse products with prepackaged solutions for Windows NT and Unix servers in response to this need. These integrated suites augment data mining tools with metadata links, the function of which is to describe the structure, attributes and changes of information in a database. They also include vertical applications such as sales force automation, call center and campaign management software, the latter being a program for sales and marketing departments that monitors transactions by setting up rules and parameters.

Informix, for example, next month will announce an upgrade to its Decision Frontier decision-support suite that adds data mining into the mix. On top of that, Informix will build four key vertical applications for the telecommunications, retail, finance and manufacturing industries. The price for such a package, say officials in Menlo Park, Calif., will fall around the $100,000 mark.

Similarly, SAS, of Cary, N.C., will round out its CRM (customer relationship management) strategy with a new prepackaged solution that ties together its SAS Warehouse Administrator Release 1.3 and Enterprise Miner 3.0 data mining tools, linked together at the metadata level. As part of its CRM solution, SAS has also set up a strategic partnership with Boston-based consultancy Exchange Applications Inc. to include links to that company's marketing automation software, called Valex. The total package will be available at the end of the year.

As a result of these and other efforts, the infrastructure will be in place to take advantage of the enormous amount of business and customer data that has been collected over the years, according to industry experts. A decade ago, such high-level, statistical functions were relegated to special government and educational projects and were dependent upon massively parallel processing supercomputers that used custom-fit neural networks and decision trees costing many millions of dollars.

"It is a top-down, strategic investment (in technology) to optimize the customer relationship," observed Henry Morris, an analyst at International Data Corp., in Framingham, Mass. "Everyone is talking about one-to-one marketing and intimate relationships, but the implication is that you are taking advantage of every opportunity to interact with the customer. That is where data mining and other types of technology come in."

At Mellon Bank Corp., efforts are well under way with this new technology to redirect the company's focus away from specific products to getting to understand and know the customer, according to Peter Johnson, vice president of strategic technology at the Pittsburgh-based bank. "The transaction processing systems were designed to look at customers with respect to product," Johnson says. "The challenge is to move the product-centric focus to look at customer relationships that will optimize service."

The shift should also facilitate designing products that better serve customers' needs, Johnson explains. For instance, consider The Lion Account, an asset management offering that Mellon acquired when it merged with Dreyfus Service Corp. in August 1994. The Lion Account integrates products ranging from checking to credit card and brokerage services, offering the customer a consolidated statement of stocks and bonds as well as tools to manage money, such as direct deposit and automatic bill paying. The account, which appeals to a broad base of financial customers, can be evolved into services tailored to smaller, more focused groups, or even individuals, by leveraging these new integrated data mining suites, according to Johnson. "This is the kind of product bundle that is the wave of the future," he says.

But to get to that point where product services are tailored to personal customer needs, Mellon first has to identify the individual requirements of the millions of customers that the bank services. To do that, Johnson has enlisted IBM's Intelligent Miner data mining tool and DB2 relational database along with Exchange Application's Valex campaign management software. The Valex software monitors the millions of customer transactions, enabling an administrator to set up rules that dictate how a customer service or sales representative should respond in accordance with an individual customer's buying and usage patterns of the bank's services. Mellon can also leverage this data to design new custom product bundles as well as enable customer service reps to cross-sell products.

This focus on CRM isn't new for Mellon. Like many other banks, Mellon has had to get aggressive in curtailing the high customer turnover rates that were the result of massive consolidation in the financial sector. To bolster its customer service tactics, Mellon began working with IBM researchers to help design a product that could do predictive analysis of customer behavior. Their efforts resulted in IBM's Intelligent Miner integrated offering.

"They weren't really inventing new technology -- all of the components existed prior to the product conception," Johnson noted. "But they integrated software components and engineering to work in a data mining environment that could mine directly against a database. It scales very well against the DB2 relational database." Pulling the components together into a GUI interface has made the data mining solution easy to use, Johnson says. But it still requires some experience and knowledge of the practices of data mining, he says.

"You need the skills of a statistician, the skills of a computer scientist to understand the structure of the database and the skills of a business analyst to frame the problem in a way that makes sense," Johnson says. "Data mining is the carrot that justifies the expense of data warehousing, yet the data mining tool is only 20 percent of the solution. The rest is your ability to apply the tool to solve the problem."

Telecommunication carriers are another industry segment looking to l everage these new data mining suites to proactively target customers with new service offerings. In their case, data mining is quickly becoming an important weapon in the vendors' fight against churn -- switching by customers to other carriers, a common occurrence in today's competitive deregulated telecommunications climate.

At Omnitel Pronto Italia S.p.A., a cellular phone company in Italy, CIO Paolo Huscher is gearing up to launch a major data mining initiative that will help the company identify customer needs. The goal is to deliver tailored customer services, such as a pay-as-you-go billing system or a different usage base plan, and thereby reduce churn.

"We want to anticipate their problems or understand their needs before it becomes a problem ... and (data mining) is the way to do it," noted Huscher, in Milan, Italy. To uphold the corporate action plan of providing top customer service, Huscher is implementing Thinking Machines Corp.'s Darwin data mining tool, which has just been put into production at Omnitel after six months of testing. Huscher is integrating Darwin with an Oracle database that houses customer billing information. Darwin, which Thinking Machines reintroduced when it emerged from Chapter 11, runs on Sun Microsystems Inc. Solaris systems, Hewlett-Packard Co. HP-UX servers and IBM AIX minicomputers. At the end of the year, Thinking Machines will port the software to Windows NT. Omnitel, for its part, runs Darwin on a Sun server with the Solaris operating system.

While Darwin licensing and support will cost Omnitel close to $350,000, Huscher said it will pay for itself by helping to stem the tide of customer defections. Omnitel supports about 4.5 million customers, any of which could defect to another cellular phone company at any time. If Omnitel can understand the parameters that signal when a customer is likely to churn, then it can combat the action with proactive marketing tactics.

And keeping existing customers is nothing short of critical to the health of Omnitel's business. "The cost to acquire a new customer is in the range of $35 per person," Huscher said, which includes sales and promotional costs. "Retaining the customer is a huge value for us -- we save this money and lower acquisition efforts."

With data mining tools, companies such as Omnitel can sift through hundreds, maybe thousands, of different variables to extract the right information from the terabytes of data in a warehouse. To the person making the query, a new pattern might be discovered -- one that may not be so obvious to the naked eye. For instance, "In the cell phone industry, they found a strong correlation between customers that purchase certain pieces of hi-fi equipment with people likely to be a regular churner," commented Charles Berger, vice president of marketing at Thinking Machines, in Burlington, Mass.

ICG NetCom Telecom Inc., an Englewood, Colo., provider of local dial-tone and IP telephony for long distance, is also concerned about tracking the right customers for promotional campaigns. "Pinpoint target marketing instead of mass marketing," which wastes money, said Badri Elosta, senior manager of database marketing and market research. "If we target a ZIP code, you might send out 30,000 mailings and hope to get the return you desire," Elosta says. But data mining tools can alleviate much of that guesswork. "In most cases, 80 percent of revenue may come from 20 percent of the customers. We want to understand these customers better."

To do that, ICG is using SAS' Data Miner with Oracle data marts and the Valex campaign management software running on a Sequent Computer Systems eight-way multiprocessing server running Sequent's Dynix/ptx operating system. ICG collects information from its own promotions, as well as buying demographic data on households and businesses from Intelliquest and Dun & Bradstreet.

With its new system, ICG marketing representatives can issue queries against all this information based on certain parameters and tailor promotions to a portion of the customer base, directing, for example, an IP telephony service exclusively to households or businesses with Internet access. "It's like paying attention to your customers...which is a win/win situation for us and them," Elosta says.

A winning situation is what Bob Roberts got when he installed a customer relationship system, based on MicroStrategy Inc.'s decision support software, called DSS Suite, and ICL Retail Systems' consumer relationship marketing solution, Corema, at Camelot Music Inc. last year. "We can make a query and slice and dice it any way we wish," said Roberts, director of new business development at Camelot Music, in Canton, Ohio. The result is information that the music company can really use to promote its products.

For instance, by querying the data warehouse, which stores information on about three million people who hold frequent shopper cards providing name, age, sex, race etc., a Camelot manager can find out what other items were typically purchased by 35-year-old women buying Celine Dion CDs. The results, Roberts says, could lead to the packaging of products in a promotion and a mailing to customers who enjoy Celine Dion's music.

More importantly, Roberts observes, is the ability to target certain customers keeps them coming back. He says people with frequent shopper cards spend 30 percent more per visit and visit 50 percent more often than other individuals whose purchases are also recorded. That, Roberts says, makes his investment in the data warehousing tools worthwhile.

Benefits aside, ICG's Elosta and others caution that integrating data mining with warehouses is not as easy as a point-and-click installation. And while these client/server applications may be available to any user in the company, the person querying the system must have some in-depth knowledge of statistics to ensure the wrong information is not extracted.

Still, the idea of establishing metadata links and bundling the three solutions, data warehouses, data mining and vertical applications -- is taking much of the headache away from the initial installation and integration. "You still have to be a trained statistician to validate results because it's easy to get in trouble with these tools," noted Bob Walters, vice president for data warehousing business development at Informix. "But when the tools are embedded for specific tasks, statisticians will be needed less and less."

Data mining suites will not be turnkey implementations, Walters says, but rather quick-start kits. However, the applications will offer an advantage in keeping customers happy. "Data mining is great for customer relationship management because it is meant to effectively combat uncertainty. And marketing data is uncertain, noisy data," Walters said.