Electronic Payment Services Inc. will move the bulk of its vast transactional data stores to a new massively parallel microprocessing system in hopes of putting the information to better use.
The company, based in Wilmington, Del., is the bank-owned parent of both Money Access Service Inc., which runs MAC, one of the nation's busiest automated teller machine networks, and Buypass Corp., a prominent point of sale processor.
In February Electronic Payment Services will begin to transfer transaction settlement data and information about their 1,500 member institutions from an International Business Machines Corp. mainframe system to a massively parallel processing system from AT&T Global Information Solutions. The new system cost $1.8 million.
Executives said the massively parallel system will allow more in-depth data analysis than the mainframe and is also cheaper to build and expand.
"Massively parallel processing is really the technology of the future," said Jeff Michel, the company's chief technology officer.
"We're able to put together these machines from commodity pieces...so economies of scale allow them to be easier to develop."
The conversion is expected to be completed by July, at which point the mainframe system will be completely cleared of its functions.
Fault-tolerant systems from Tandem Computers Inc. will continue to handle transaction traffic for Buypass and MAC.
Massively parallel systems typically consist of loosely-knit clusters of microprocessors linked together to break down and process information concurrently rather than successively.
This much ballyhooed computing architecture -- considered ideal for data base management, or "data mining" applications -- has recently been adopted by many financial institutions.
Banks and related companies are employing massively parallel processing primarily to segment customer groups.
With customers grouped with like customers, banks can tailor their marketing and identify cross-selling opportunities more readily.
In addition, segmentation allows institutions to identify their best customers and to work up products to satisfy them.
Electronic Payment Services expects to do much the same with its own system -- giving itself and its bank customers a more focused picture of the customers who use their ATMs, according to Mr. Michel.
Another benefit of parallel processing is the scalability that it brings to an institution or network computer system.
With a mainframe, Mr. Michel said, the company was forced to "upgrade in big chunks." Whereas, with a parallel computing system, it can expand in "very finite chunks."
"If we want to upgrade every quarter, we can do that," he said.
The new system will consist of 12 Pentium microprocessors with 2 gigabytes -- or 2 billion bytes -- of space.
But the company plans to add another six processors soon, observers said.
For now though, the payment processor will have its hands full with the monumental task of converting the billing system, settlement and reporting data, and participant files pertaining to all of its 1,500 bank members.
The company processes more than one billion transactions a year through the MAC network and more than 600 million through Buypass.
Throughout next year, executives plan to perform some extensive data mining to help them better understand their bank members and their customers.
The company has about a dozen initiatives in the works, including plans to research patterns of ATM usage in different regions, according to Mr. Michel.
By 1996, executives anticipate extending these information services to the company's financial institution members to aid them in better understanding the habits of their consumers.
"We need the effectiveness of mass marketing, but we also need to treat people as individuals," Mr. Michel said. "This technology is allowing business to be conducted differently than in the past."
Data to Be Handled By New System
Settlement and reporting
data for 1.65 billion annual
ATM and POS transactions
for 1,500 member
Participant files on
Source: Electronic Payment Services Inc.