Developed non-traditional credit scoring algorithm for a major global bank to simplify approval process for 50% of credit card applications

Financial Services, Risk Management

We simplified the credit approval process for a major global bank by developing a credit policy based on traditional & non-traditional data. The central pillar of the approach was an ensemble model that predicted an applicant’s income, using traditional credit bureau data, along with alternative information, such as geography, demographics, employment information, etc. The developed policy allowed 50% of applicants to go through real-time approvals, thereby reducing leakage and negative selection.


  • Indian credit card business relied heavily on physical document collection.
  • Customers had to submit detailed income documents, leading to lost sales, negative selection, and inefficient process.
  • Used combination of employment, geographical, and bureau information to predict actual incomes.
  • Tested various modeling approaches




  • Designed a strategy that balanced potential risk against operational efficiency.
  • Strategy allowed >50% of applicants to go through the process without income document collection.
  • Designed a testing and measurement plan to measure risk impact.