Situation
Client was experiencing high delinquency rates on its New to Bank Credit Card bookings.
Approach
Axtria built a predictive model to modify the acquisition strategy and create an initial line assignment strategy.
- Data Collection
- Collected approved accounts data for last 24 Months
- Collected Bureau data
- Collected Application data
- Created data dictionary to identify missing rates and reason for missing values
- Outlier identification
- Data Collection
- Created definition of default as 6M 60 DPD rate
- Independent variable creation: bureau score, bureau variables, application variables
- Missing value imputation
- Outlier treatment
- Variable reduction using various techniques:
− CART
− One step logistic
− Bivariate plots
- Segmentation
- Created segmentation to identify various segments of accounts
- Identification of segments:
− Red: Default rate>20%
− Amber: Default rate 10-20%
− Green: Default rate<5%
- Computed 12M EBIT for each segment thus created
- Swap out sets identified with 12M EBIT as negative
- Validation of segmentation on OOT samples
- Line Assignment Strategy
- Average 12 M Utilization look up tables created by credit limit
- For each segment sensitivity of EBIT computed for various combinations of credit limit – utilization
Result
- Created as segmentation for identifying swap out sets
- Positive EBIT impact on Swap Out
- Created a line assignment strategy
- New Strategy resulted in $15 Mn savings in NCL