HSBC Takes Scientific Approach to Lending

A streamlined installment loan decision process has allowed HSBC Holdings to strengthen its balance sheet by extending credit to customers with the highest expected profitability, not just those with the lowest default risk.

Using the Strategy Science optimization methodology from Fair, Isaac, London-based HSBC has built a statistical model to optimize the installment loan process.

"No one had done it before," said George Lennox, senior manager for group credit and risk at $696 billion HSBC Holdings plc. "We were asking them Fair, Isaac to transfer the technique that they had developed for cards, and to see whether it worked on loans."

The bank was convinced by the results of a "swap set" analysis comparing its previous decision-making process to the one developed through Strategy Science. "At this stage, they proved to us that they can outperform any strategies that we have, and therefore it's worth investing in," said Lennox. "We'd be silly to ignore it."

The Strategy Science technique involves a statistical method known as Bayesian network learning, which is driven by "influence diagrams" that model the actual decision process.

But Strategy Science is more than just a computer program. Building a valid influence model requires hands-on participation from both business analysts and operations researchers, along with access to a diverse repository of actual data.

"They're building these equations and testing these relationships one at a time," said Lori Sheerer, vice president at San Rafael, Calif.-based Fair, Isaac. "There's a huge experience curve, because the development of these decision models is a hypothesis-generated approach."

HSBC analysts worked closely with Fair, Isaac to tweak the Strategy Science model and to backtest the results against past installment loan data.

"Both parties, HSBC and Fair, Isaac, had to get into research mode and do quite a bit of work before we had something that was viable," said Lennox.

HSBC has been rewarded with greater internal confidence in its lending decisions. For its part, Fair, Isaac can apply what it learns from the HSBC experience about installment loans with future clients.

Now that HSBC will be using Fair, Isaac for both credit card and installment loans, the bank has begun to consider a strategy for managing credit exposure at the customer relationship level across the entire brand.

"A nice situation would be that the customers know they can always log on and see their line of credit," said Lennox. "Whenever they need it, they can book it into a particular mechanism or product."

The plan is to create a common interface to a range of lending products typically managed by different business units or by monoline issuers.

"Let's say we had the line of credit calculation that arrived at a figure of $30,000," said Lennox. "Do you want it to be a revolving line of credit? Do you want it in an overdraft on a demand basis? Do you want to own it on an installment basis? You decide what's best for you."

Bank customers could reduce credit lines on demand, thus freeing up the institution's contingent liability for more profitable financial purposes.

"They can use what limit they like, in what manner they like, wherever they like," said Lennox.

To be sure, there are some barriers, such as limited access to e-banking among the general population and the need to link together several repositories of customer and loan information.

"But if we collectively put our minds to it," said Lennox, "we could be talking about something that's available soon."