Quadstone Brings Data Mining Offerings U.S. on NT

If most sales leads hang up on your call center telemarketers, a new combination of tools from a Scottish data mining outfit could help with that and similar problems.

Quadstone Ltd. (www.quadstone.com) announced the port of its Decisionhouse data mining tool to Windows NT earlier this year. The product has been sold mostly in Europe on Unix since 1996.

In April the company made an add-on to Decisionhouse, called Transactionhouse, which is available on Unix platforms. An NT version is expected to be released in July. Transactionhouse is designed to loop the results of data mining analysis back into a company’s operational systems.

For example, the combination on NT allows a business analyst to use the results of recent telemarketing efforts to refine the definition of a customer who is a likely sale. Transactionhouse rolls transaction data from the call center or other transaction records back into the company’s customer database, up to 10 million names. With Decisionhouse, a business analyst can re-examine the data for new or undiscovered patterns and can rescore the database based on the new criteria. Transactionhouse then comes in again, taking the new data and plugging the top 100,000 or so names into telemarketers’ calling queues.

"When you’ve called the first 10,000, you can analyze what's working," says Mark Smith, Quadstone’s vice president for marketing. "Re-evaluate the whole thing. Rescore all 10 million. It’s being done right on the back of real attempts to sell products."

Smith says e-commerce is ratcheting up the speed of the sales cycle and, therefore, the need for immediate analysis and response. "People are now able to change the offer they’re making to customers within hours by showing it on the Web site."

Quadstone opened a U.S. office in Boston last year and is making a consolidated push into the U.S. market this year. Most of its customers are in Europe.

Quadstone officials say the features that put their product in a class of its own include a tight focus on customer segmentation and analysis, ease of use for business users without statistical backgrounds and strong parallelization.

"Most of the tools are actually to be run by people in the IT department who don’t have this business view. The marketers … need to get stuff done. [They can’t] wait two weeks for some answer to come back from the stats guy," Smith says. "We’re a tool for them to make that best guess."

Herb Edelstein, president of the data mining consulting company Two Crows Corp. (www.twocrows.com), takes a skeptical view of vendor claims of having a focus so narrow that there are few or no direct competitors, Quadstone’s included. Most data mining products are targeted at customer modeling in some form and claim ease of use, Edelstein says.

On Windows NT, Edelstein says the company will be in tight competition with IBM Corp.’s Intelligent Miner, SAS’s Enterprise Miner, SPSS AnswerTree, Integral Solution’s Clementine and two soon-to-be-released-on-NT tools: Silicon Graphics Inc.’s MineSet and Thinking Machine’s Darwin.

"I think there are products like Model 1 from Group 1 ... which, at the sacrifice of some flexibility which Quadstone has, nevertheless make things available to less-skilled business users," Edelstein explains. "There’s a spectrum of business users."

On the other hand, Quadstone has a number of features targeted directly for customer analysis and segmentation, Edelstein notes. They include a user-friendly interface, solid data preparation tools, strong visualization, batch processing and a heavy focus on scoring. "It’s very clear, if you look at their history, that this is very much where they want to keep their focus," Edelstein says. "It’s a very respectable product."

Many of the scalability benefits Quadstone gets from parallel processing don’t apply yet to Windows NT. Quadstone has scored a 100-million-customer list on an eight-way Sun Solaris system. But the company only recommends Windows NT up to about 10 million customer records. That number is cut to about 5 million for companies that expect to put two to four concurrent users on the system.

"At the moment, we don’t have a whole lot of experience on NT above the middle tier," says Patrick Surry, Quadstone’s vice president of services. The company is working with Microsoft Corp. and Data General Corp. (www.dg.com) to scale up on the platform, Smith says.

Both Decisionhouse and Transactionhouse will run on a dedicated machine Quadstone calls an analytical server machine. Decisionhouse uses a three-tier architecture, which accesses a data source, pulls the necessary data into a virtual memory management system designed for parallel processing on the middle tier and shows results to the client on another tier. The top of the Windows NT chain with Quadstone consists of a four-way machine with a couple gigabytes of memory and up to 50 GB of storage, Surry says.

The company’s tools have native access to Oracle, NCR Teradata and Microsoft SQL Server databases and use ODBC to access others, Surry says.