HSBC tackles credit card fraud with new system

Business analytics specialist SAS today announced a new fraud management platform, developed in partnership with financial services giant HSBC.

The fraud management tools provide a real time card fraud detection system that works faster, finds more potential fraud with fewer false positives, and needs fewer IT resources. HSBC has already implemented the service in the US to protect 30 million cards, and will be deploying it in the UK for credit card fraud management in the second quarter of 2008, followed by a deployment for debit card use in the third quarter of 2008.

Card fraud is big business. In the UK the estimated losses to card fraud topped 400 million in 2006, with about 100 million coming from counterfeit cards. HSBC wanted to replace its existing Fair Isaacs fraud detection system with a platform it could build on to bring all its fraud detection into one place. Derek Wylde, head of group fraud at HSBC, was keen to consolidate. "With a business the size of HSBC we do have a number of different systems around the group. It is incredibly difficult and time consuming to change solutions too often; we want to make a commitment with a system that is future-proof and will last into the next 15 to 20 years."

The real time card fraud detection system SAS developed with HSBC can decide whether to approve a transaction in 30-50ms. Built using SAS's existing risk management and analytics tools, the system uses a mainframe scoring agent to look for patterns that indicate fraud, with a neural network-based model to assess the likelihood of a transaction being fraudulent. Reporting tools give HSBC the opportunity to test and modify rules to detect more fraud with fewer false positives.

While HSBC isn't giving specific figures for improvements in its own fraud detection, the SAS system is claimed to be more efficient; it's dealing with 87 per cent more data with 12 per cent less mainframe processing overhead which halves the mainframe processing cost per data item. That reduces the cost of detecting fraud considerably, with a 30 per cent decrease in the computing resource cost of processing potentially fraudulent transactions, and a 10 per cent increase in efficiency for agents investigating those cases. So far, the new platform has replaced three separate applications which has also reduced IT support costs.

Wylde plans to replace more existing systems and improve fraud detection by monitoring transactions and other customer behaviour - like a fraudster trying to change the address for an account. "Cheques, internet payments, telephone payments; all these potential transactions have fraud in them. We've already begun exploring with SAS and our Hong Kong business how we can deploy the system to monitor and detect fraud for more than payment cards - for cheques, for bill payments, Internet payments, non-monetary transactions and potentially even for anomalous staff behaviour. We could use it for credit risk management, detecting early signs of delinquency and bankruptcy. There's even something for the marketing guys; using transaction data can be incredibly valuable when pinpointing where to sell products to our customers and the SAS solution can be used for that as well."

He sees the system as a way of avoiding the cost of issuing tokens to banking customers but it could also enable micro payments and low value transactions on contactless cards. "The cost of processing is coming down all the time not just in terms of computer processing but in telecoms costs. It's not just that the cost of fraud detection is more cost efficient; the whole infrastructure underlying card payments is becoming more cost efficient. The cost of getting an authorisation from the point of sale to the host system is cheaper than it was five or 10 years ago."