Howe Faculty Receives Patent to Help Financial Services Companies Better Identify Customers

Professor Germán Creamer Received Patent


In the aftermath of the 2008 financial crisis, many firms found themselves picking up the pieces from clients who should have been flagged as at-risk. But, as many firms discovered, identifying these individuals among other similar customers was a difficult undertaking. The problem is known as entity resolution and it has become an increased opportunity under enterprise-wide risk management systems. Firms began to recognize that their information systems might not match customers with their account information and records as thoroughly as possible. But thanks to a patent process that had begun in 2006, Professor Germán Creamer, Associate Professor of Quantitative Finance, set out to help firms better identify those customers. At the time, Creamer was a full-time staff member of the Information Management Group of American Express. “The company, as with many Fortune 500 companies, outsourced work like this,” says Creamer. “When I joined the analytical team, however, I saw a problem of how to identify customers. With entity resolution there could be a person John B. Smith but he would be recognized as John Smith. I thought, ‘how can we assign that person’s name or account to one specific person using the existent customers’ information?’ It was interesting because the approach applied computer science methods to solve a customer management problem. Large corporations had a problem because they had very large databases to identify their customers without the correct information for outreach.” Creamer saw a way to use databases better. First, he would need to create a methodology to generate a matching accuracy score to determine the specific person. Then the matching accuracy score could be analyzed with the external database of credit risk scores to identify their level of risk. Afterward the person’s identity and accuracy score could be analyzed with risk-analysis variables to assess any hidden risks.

“What I proposed was a machine learning algorithm that would consolidate the information and create a score – one being a low level of accuracy and five an exact match – to indicate how accurate the identification with the customer was. Once we had the match we could implement the second level and study the relationship between matching accuracy and risk segments to minimize the exposure to risky customers. We wanted to be able to reach only good customers. We don’t want to target customers that may buy certain products but don’t pay.” Creamer’s approach paid off. The process began in 2006 and a patent was filed in 2008. But it was only in November of 2011 that the patent was issued.

When Creamer joined Stevens in 2009 he applied similar thinking in preparing the proposal for the Hanlon Financial Systems Lab. This is a state of the art finance lab that opened in April of 2012 and is dedicated to study and replicate the complexity of financial markets. “Most of the problems that financial companies face today use some sort of information technology approach,” Creamer says. “They could not manage their business without the capacity to extract information and generate new strategies from large corporate databases.”


  • German Creamer - New Jersey
  • Prashant Sharad Churi - Arizona
  • Sara Tresh – New York
  • Mary Weissman - Arizona

For more information please email Professor Creamer