Pattern recognition software and dramas of deception: new challenges in electronic financial services

RMA Journal, The, Oct, 2004 by James F. Bauerle

Patterns of deception are built on drama. Like magicians at a circus, those practicing the deception must divert attention from what is actually taking place by creating another story that masks reality. We may like to refer to these people as "bad actors," but the successful ones are very good actors indeed. Six tips are provided to help the banker make these actors take their final bow.

Pattern recognition software is one of the most promising growth trends in financial services in this decade. Today's reality is that pattern recognition software is everywhere. A species of artificial intelligence, it influences business and consumer behavior in ways that most people do not the take time to recognize, much less analyze. amazon.com uses software to identify the buying habits of its customers and present them with items they are likely to want based on their buying history. Credit-scoring software has become so sophisticated that risky customers can be spotted based on behaviors, like poor driving records, that superficially have nothing to do with credit management. Credit card issuers, engaged in a war of attrition for good customers, fling computer-generated, attractively priced loan offers at customers whose payment histories show large monthly payoffs. So where does all this end? Is it a net benefit to financial services companies and their customers? What are the pitfalls for the unwary?

Foundations of Growth

The proliferation of pattern recognition software (PRS) is rooted in several continuing developments. Most fundamentally, continuous advances in computer technology make it possible to write and execute algorithms that identify important patterns across large populations of data and that respond to the observed patterns in ways that promote desired behaviors and discourage undesirable ones. Customer relationship management software is a leading example.

Supporting the development of this technology are recent legal changes that encourage and protect it. From the 1960s to the late 1970s, the U.S. Supreme Court said software could not be protected by patent. Software, the reasoning went, was merely the expression of mathematical formulas. As such, anyone could create or recreate them, and patent protection should not be available. Beginning in the early 1990s, the U.S. Court of Appeals for the Federal Circuit began to take a different tack. "The proper inquiry in dealing with the so-called mathematical subject matter exception is to see whether the claimed subject matter as a whole is a disembodied mathematical concept ...," the court wrote in a 1994 case. (1) Four years later, in the State Street Bank case, the court made clear its intention to reverse decades of settled case law. The court reasoned that the software in question no longer represented only mathematical processes, but constituted a machine when operated in tandem with associated hardware. As a machine, the software was an invention under federal patent law. (2)

The result of these developments has been an explosion in the number of patents granted in patent class 705--financial, business practice, management, or cost/price determination (data-processing) patents. During the early 1980s, the U.S. Patent and Trademark Office issued an annual average of 60 patents in this class. During the early 1990s, the annual average rose to 208. By 2001, the last year for which data was published, the annual average reached 980, equivalent to all the patents of this class issued during the 1980s! Equally significant, leading financial services companies now join technology companies as recipients of the largest numbers of these patents.

PRS Classified

As the class of patented software has expanded, so have its uses. Pattern recognition software, in particular, can be classified into at least four categories, according to the pattern that the software works to deconstruct.

Patterns of obligation uses software to measure how well or poorly people meet their obligations. Credit-scoring software is the leading example. Patterns of occupation automate banking processes previously performed by humans. This class holds huge potential for transforming the industry. Examples of processes being automated include the entire lending process, from application through administration, the process of syndicating loans, the process of selling securities to the public, and any number of other labor-intensive endeavors associated with financial services.

Patterns of desire are the focus of customer relationship management software that retailers like Amazon.com use. Rather than ask the open-ended question embodied in Microsoft Corporation's trademarked phrase, "Where would you like to go today?" this software builds on the enterprise's knowledge of the customer's buying habits and drives the customer to repeat transactions in an area that the software recognizes as a sweet spot for that customer. If a customer's data file shows that he or she typically finances a European vacation every three years with a $10,000 installment loan, why should the software not invite the customer to increase the frequency of vacations to every two years if the monthly payment to retire the associated debt can be kept nearly the same? Conversely, if a cohort of customers can be statistically proven to be indifferent to fluctuation in deposit account interest rates, why should their bank not use that information in its asset/liability model to support the purchase of longer-bond maturities in the bank's investment portfolio?