Preventing loans from default
8 Jan 2023
Overview
When loans fall in default, workout strategies are based on huge amounts of intuition work that generate a certain return
Consumer loans have an increase complexity, as low tickets lead to high volumes
This high volume – low ticket combination makes granular action impossible for manual work based structures
The Challenge
Data quality was low, and strong data engineering efforts were needed
The portfolio was composed by personal loans, credit cards and overdrafts
The Approach
Payment anticipation model: we trained a model capable of identifying paying customers before any loan management was performed
Allocation algorithm: we created a loan allocation algorithm to sort the assets between the available resources, to increase management intensity (with the same FTE), while reducing collection costs
The Results
For the whole portfolio: High score loans held a collection rate of 6.12%, while low score held a collection rate of 1.64%
On new entries (+90DPD): High score loans held a 24.16% collection rate, while low score held a 10.95% collection rate
Preventing loans from default
8 Jan 2023
Overview
When loans fall in default, workout strategies are based on huge amounts of intuition work that generate a certain return
Consumer loans have an increase complexity, as low tickets lead to high volumes
This high volume – low ticket combination makes granular action impossible for manual work based structures
The Challenge
Data quality was low, and strong data engineering efforts were needed
The portfolio was composed by personal loans, credit cards and overdrafts
The Approach
Payment anticipation model: we trained a model capable of identifying paying customers before any loan management was performed
Allocation algorithm: we created a loan allocation algorithm to sort the assets between the available resources, to increase management intensity (with the same FTE), while reducing collection costs
The Results
For the whole portfolio: High score loans held a collection rate of 6.12%, while low score held a collection rate of 1.64%
On new entries (+90DPD): High score loans held a 24.16% collection rate, while low score held a 10.95% collection rate
Preventing loans from default
8 Jan 2023
Overview
When loans fall in default, workout strategies are based on huge amounts of intuition work that generate a certain return
Consumer loans have an increase complexity, as low tickets lead to high volumes
This high volume – low ticket combination makes granular action impossible for manual work based structures
The Challenge
Data quality was low, and strong data engineering efforts were needed
The portfolio was composed by personal loans, credit cards and overdrafts
The Approach
Payment anticipation model: we trained a model capable of identifying paying customers before any loan management was performed
Allocation algorithm: we created a loan allocation algorithm to sort the assets between the available resources, to increase management intensity (with the same FTE), while reducing collection costs
The Results
For the whole portfolio: High score loans held a collection rate of 6.12%, while low score held a collection rate of 1.64%
On new entries (+90DPD): High score loans held a 24.16% collection rate, while low score held a 10.95% collection rate