International Business Machines Corp. has been hired to analyze a massive pool of operational risk data that the Operational Riskdata Exchange Association, a consortium of 37 major international banking companies, has been compiling for years.
The Armonk, N.Y., company signed a contract last week to look for trends and patterns in data contained in the database.
The consortium of North American and European banking companies created the database in 2002 to pool information anonymously. Members began populating the database in 2004, but the data has been largely untapped.
"It's the first time that this type of capability is going to be dedicated to the data set that ORX has," said Joe Sabatini, the chairman of the consortium and the managing director and head of corporate operational risk for JPMorgan Chase & Co. The information within "is a window on the industry" and could provide some powerful insights for the banks that have access to it.
Today there are about 63,000 entries in the database, he said.
IBM is absolutely necessary to this process, Mr. Sabatini said. Any company could perform its own analysis on its own data, but "the difficulty is that no individual member has the analytic resource" to "do analysis across the full pool of data."
Because the entries are meant to be anonymous, some important details have been stripped out of them for privacy concerns, he said. For example, information such as the geographic region in which a loss occurred, or the revenue for the business line that incurred a loss, is taken out.
However, as a third party, IBM will have access to some of that information to conduct more thorough research. It will not include the redacted information in its final results.
"IBM will have the data, but they won't have the name of the banks associated with the individual data points," Mr. Sabatini said. "But they have data that no individual member has."
The database's contents are "robust enough to devote some high-powered analytics to it."
IBM expects to begin its work this month, though there is no specific schedule, or even a firm description, for the project yet. "We're in the very early stages," Mr. Sabatini said, and the consortium's members do not yet know "what sort of analysis will be insightful, beneficial, and valuable to the members that IBM can produce."
June Yee Felix, the general manager of banking solutions and strategy for IBM, said that the project probably will take about three years, and that her company hopes to provide analysis that can help bankers adjust their business decisions by determining how their losses and spending compare with the industry's.
For example, a company can compare its losses in one business line to the industry average for a company its size and then make informed budget decisions, she said. "This is going to be a big transformation for the industry."
Cubillas Ding, a senior analyst with the Boston market research company Celent LLC, wrote in an e-mail that many financial companies think that their data for modeling risk at this level "is either incomplete, unavailable, or lacks relevance for their firm's business." Using information from a comprehensive database to "supplement internal databases" will help companies "prioritize their performance improvement efforts."
However, he also warned that bankers need to be wary of looking at data that may lack context. "Sometimes drawing conclusions from benchmarking studies without fully considering the circumstances" surrounding a loss "can lead to misleading deductions," he said.
