CIBC uses analytics to put customers under microscope

The Canadian Imperial Bank of Commerce is using scorecard technology to analyze credit risk for its more than nine million customers.

The North American financial institution has rolled out SAS Credit Scoring to produce credit risk scorecards,

as well as perform analytics on its retail portfolio. The latest version of SAS Credit Scoring helps banks develop scorecards to evaluate credit worthiness – and help minimize business risks. It provides a basis for scoring loan applicants for credit cards, personal loans and mortgages, among others.

“We are currently using the credit scoring solution to aid in the development of both application and behavioural scorecards for our retail portfolio,” said Sanjiv Talwar, vice-president of retail risk management with the CIBC.

The CIBC updated its entire retail analytical modeling environment at the beginning of the year, which included the introduction of the credit scoring application. The tool is now available to all modellers working on credit risk solutions for the bank’s retail portfolio.

Banks use selected information about an applicant – such as how much money they earn or how long they’ve been working – in a scorecard to come up with a probability of default.

Naeem Siddiqi, credit scoring strategist with SAS, said the application helps users to develop these scorecards in-house. Its original application is now integrated with SAS Credit Risk Management, which performs credit data management, credit scoring, credit portfolio risk management, regulatory risk-weighed assets and capital adequacy calculations.
“You can develop the models that will help you predict the probability of default, then feed the results of these models into the larger solution which you can then use to calculate your risk and ultimately your economic and regulatory capital,” he said.

Talwar said he sees benefits to developing an analytical capability within the CIBC, such as providing greater insights into customer behaviour. For example, it can segment retail portfolios into “homogenous pools,” which means it can group accounts with behavioural characteristics to show consistent behaviour over time.

But, as with any new tool, there is a learning curve. “We have a number of sources to turn to for help, be it from SAS, to co-workers who may have used the tool, to online forums,” he said.

Each modelling project is aimed at solving a particular business issue, and each issue has associated costs. “As part of our project delivery, we attempt to quantify the benefits associated with the modelling solution,” he said. Some of these benefits include increased efficiency, reduced errors and standardized methodology, he said, but determining exactly what percentage is difficult.

Traditionally it took banks between four to six months to develop a scorecard, but now scorecards can be produced as quickly as two weeks to a month and a half, said Siddiqi.

“This solution helps banks reduce their losses,” he said. “It directly affects the bottom line. The ROI is very direct. If you develop more than four scorecards a year, it’s cheaper to do it in-house. A large bank may be developing between 15 and 30 scorecards every year so for them doing it in-house has serious economies of scale.” Most customers recoup their investment in six to 12 months, he added.

At the CIBC, credit scorecards are recalibrated on an as-needed basis, said Talwar. “We currently review our scorecards for performance deterioration or significant population shift on a regular basis,” he said, adding this is used to help decide whether to build a new scorecard. “It is difficult to say what the average number of scorecards per year is because it is really a needs-based result.”

Comment: info@itbusiness.ca

Share on LinkedIn Share with Google+
More Articles