Business Summary Report:
Predictive Insights for Collections
Strategy
1. Summary of Predictive Insights
Our predictive model identified several customer segments at elevated risk of credit card
delinquency. Key risk indicators include high credit utilization, missed payments, and
elevated debt-to-income ratios. These insights can help prioritize which customers may
benefit most from early outreach or financial support strategies.
Key Insights Summary Table:
Key Insight Customer Segment Influencing Potential Impact
Variables
High credit Customers with Credit Utilization, Consider lowering
utilization >50% utilization Missed Payments credit limits or
correlates with offering usage
increased monitoring tools.
delinquency risk.
Young customers Under 30, 2+ Age, Payment Proactive outreach
with missed missed payments History with tailored
payments are high financial education
risk. or hardship
support.
High DTI is DTI > 0.5 Debt-to-Income Debt restructuring
associated with Ratio support or
higher default repayment plan
rates. options.
2. Recommendation Framework
Restated Insight:
Customers under 30 with two or more missed payments have a significantly higher
likelihood of delinquency.
Proposed Recommendation:
Launch a 6-week pilot outreach campaign targeting this segment with proactive SMS
and email messaging. The goal is to offer tailored payment plans or financial counselling
support before accounts reach 30+ days delinquent.
Justification and Business Rationale:
Specific: Focused on a clearly defined, high-risk group.
Measurable: Target a 10–15% reduction in delinquency within the pilot group.
Actionable: Uses existing communication infrastructure.
Relevant: Aligns with Geldium’s goals to reduce credit risk and improve customer
outcomes.
Time-bound: Designed as a time-limited pilot with measurable outcomes.
3. Ethical and Responsible AI Considerations
The model was evaluated for fairness using multiple performance metrics across age
and income groups. No disproportionate flagging of protected segments was observed.
Bias: We tested for overrepresentation in delinquency predictions and found a
balanced outcome across customer demographics.
Explainability: The model uses logistic regression, which allows clear explanation
of how key variables influence predictions.
Responsible use: The recommendation is focused on early, supportive
interventions rather than punitive action, reinforcing fairness and customer care
principles.
