Maximizing Credit Card Profitability with ML–Driven CLV

on YouTube ↗34m

" summary

Uses AmEx experience and peer-bank practice to build CLV that predicts future spend, revolving amount, payment reliability, and default probability and drives actions. Models are predominantly XGBoost/LightGBM with optional deep‑learning sequences and survival time‑to‑event layers. Validation requires temporal validation, segment fit checks, test/control pilots, quarterly monitoring and stress factors. Claims include shortening typical 4–6 year breakeven to under four years for targeted cohorts, enabling higher sign‑on offers, dynamic credit‑line adjustments, and measurable improvement in risk/reward and resource allocation.

AI-written summary · about

" chapters (12)

1. 0:00Welcome & Talk Topic: ML-Driven Customer Lifetime Value
2. 1:20Agenda Overview: Credit Card Dynamics, CLV, ML, and Use Cases
3. 2:08Credit Card Economics 101: Acquisition vs. Customer Management
4. 4:45Why Customer Lifetime Value Matters (Acquisition, Portfolio, Forecasting)
5. 7:14Machine Learning as the CLV Game Changer
6. 8:56Key CLV Model Outputs: Spend, Revolve, Pay, Default Risk
7. 13:01What Data Powers CLV Models: Spend, Payments, Utilization, Bureau
8. 15:13From Raw Data to Models: Feature Engineering & ML Approaches
9. 18:00Model Governance: Validation, Segment Fit, Testing, Monitoring, Stress
10. 22:34Real-World Applications Across the Lifecycle (Acquisition & Management)
11. 28:05Measuring Impact: Faster Breakeven, Better Risk/Reward, Smarter Resourcing
12. 30:34Key Takeaways, Validation Warning, and Closing

" related talks

• 36mCustomer Lifetime Value: Models, Metrics and a Multitude of UsesBrian Bloniarz · PyData Los Angeles 2018
• 39mMachine Learning in Financial Credit Risk AssessmentSoledad Galli · PyData London 2017
• 24mUnlocking revenue: newcomer classification for boosted profitsDmitrii Matveev · DSC Europe 23
• 44mEating Humble Py: From Toy Problem to Real-World Solution in Predicting CLVKatherine Munro · PyData Yerevan 2022
• 32mMaking Models Work: How an optimized ML lifecycle can drive better ROIPriyank Patel · TWIMLcon: AI Platforms 2021
• 28mRecent approaches in retail banking marketing optimizationMichal Ambler · DSC Europe 23
• 45mKnow Your CustomerFiona Li · Data Con LA 2019
• 58mNew-type Scoring Models: Analyzing Customer ActionsMaksim Savchenko · РИТ++ 2017