Forward-looking economic calibrations of risk, such as those I described in my last post, are being folded into a wide range of customer decision strategies. Samsung Card, one of Korea's largest credit card companies, is at the forefront of this trend.
We recently worked with Samsung Card to improve portfolio profitability by encouraging cardholders to make greater use of an installment loan product attached to their credit card account. This product has a high margin, but currently low usage. To understand how to boost balances in the most profitable part of its portfolio while carefully controlling risk and meeting regulatory requirements under changing economic conditions, the lender is using action-effect modeling.
Samsung Card wants to know, for example, which customers are likely to increase their utilization of the loan product if offered a 30% temporary pricing discount—and which customers would be sufficiently incented with a just a 20% or even 10% offer. Action-effect models help solve such complex problems by predicting how customers will respond to various actions the issuer might take.
But lenders need to be able to understand response in the context of outcome metrics (profit, loss, exposure, cost, etc.) and have a way to translate that understanding into action. For this project, that means making offers with the most profitable combination of loan limits and pricing for each customer.
That's where we see the full power of this analytic technique. In action-effect modeling, we find mathematical relationships between lender actions, predicted customer responses and outcomes. Moreover, in this case, since the risk prediction component is economically calibrated, the decision strategies we build from the model are also forward-looking. They reflect the economic conditions the lender expects in the coming year, rather than for the conditions of past years (i.e., the last time they refreshed their risk models).
Decision strategies are optimized for maximum profit against portfolio-level constraints (losses, balance loss ratio, marketing spend, volume targets, etc.). This approach essentially provides an "interactive map" of the problem being solved. Using simulation, Samsung Card can move the decision levers—adjusting limits, prices and constraints—and the optimal decision strategy will move accordingly to a new coordinate position on the "map." Trying out "what if?" scenarios, the company can explore the trade-offs between risk and reward, and generate a range of potential optimal operating points for consideration.
The way this approach reveals the most powerful profit drivers and makes the trade-offs explicit is providing Samsung Card with business value over and above the expected profit lift. For the first time, the company's risk and marketing groups are coming together to make decisions jointly, instead of separately and sequentially, and to choose the strategies that make the most sense for their business at this time.
As the optimized strategies are deployed, and production results are collected, these groups will be able to efficiently evaluate what is working and what could be improved. Together they can drive rapid test-and-learn cycles that deliver performance improvements in the midst of economic change.