R&C: In what ways has credit scoring changed in recent years? What factors now drive the decision-management processes of lending organisations?

Siddiqi: Over the last 15 to 20 years, credit scoring has become more analytical and mathematical. There used to be many business analysts building models, but now it has become more specialised, with decision scientists taking on this responsibility. This has partly been driven by a high demand for credit scoring and modellers globally. Of course, there has also been the Basel Accords (BASEL II) and the IFRS 9 International Financial Reporting Standard, as well as various other regulations that prompted many banks, even smaller ones, to develop their own models internally.

Filipenkov: Credit scoring has dramatically changed in recent years. The availability of new data sources significantly increased the capabilities of the machine learning (ML) techniques that financial institutions can use in the lending process. The decision process is on the one hand becoming much more complex, involving more models, and on the other has become much faster due to market demand and the increasing accuracy of automated decisions. In addition, decision management and model risk management has become more complex, requiring an agile approach. New risk or marketing models are being introduced and modified every day. There may be thousands of different decision paths depending on customer segmentation to deliver customer-focused offers. This means that the model and decision management infrastructure should support this complex environment, and allow rapid changes, testing and deployment of new models and decision logic.

Jan-Mar 2019 Issue

SAS Institute Inc.