Data Science and Credit Modeling
Improving a Pricing System for a Retail Bank
Context
A bank with more than 5 million customers had been working with an ‘intuitively-developed’ policy to control important processes related to a Fee Waiver Incentive System (FWIS). It rewarded clients solely according to their relationship with the institution. Our client desired to evaluate how effective their system was and engineer a new solution based on a statistically computed model using their client's database.
Method
Machine Learning techniques like clustering, factorial analysis and different types of regressions were used to separate a Big Database with clients into strategic groups, based on socioeconomic and transactional information. Event-studies were carried out to measure the models’ performance before and after pricing discounting events.
Benefits Achieved
The client has been able to ally its business intuition to a way more precise quantitative model: a new FWIS which is enhancing the bank’s revenues as well as creating measurable loyalty from its clients.