Enhancing Non-Performing Loan Allocation with AI: Reducing cost-to-collect by +40% while increasing collection rates
Menhir Technologies Case Study for Financial Services
For more details, get in touch with our Financial Services practice at fs@menhir.ai
Introduction
The financial management landscape is ever-evolving, with non-performing loans (NPLs) posing significant challenges for banks and financial institutions. These loans, which are in default or close to being in default, can affect the health of financial institutions and economies. To better manage and allocate NPLs, institutions are turning to the potential of artificial intelligence systems (AI).
The Challenge of NPLs
NPLs arise when borrowers haven’t made scheduled payments for a specified period. Accumulation of NPLs can lead to liquidity problems for banks, making it essential to robustly allocate and manage these loans.
The Role of AI in NPL Management
Traditionally, banks relied on human expertise and basic tools. However, growing data volumes and financial complexities are making manual processes insufficient.
AI transforms NPL management through:
- Data-Driven Insights: Machine learning models predict potential loan defaults using historical data, including economic indicators.
- Allocation Optimization: Advanced AI algorithms suggest optimal routes based on predicted recovery rates and costs involved.
- Continuous Learning: AI models adapt to new data and shifting market dynamics.
But… what about explainability?
Banking regulations enforce Banks to explain AI models to avoid an kind of bias the models may add to the decision making process. This is where Explainable AI and Causal-AI comes in handy. At Menhir.ai we have developed since our inception complex explainability systems that provide simple answers to the “why is the model deciding this?”, enabling “what-if” analysis and scenario generation.
AI in Financial Services makes no sense without:
- Perfectly Engineered Models: For understanding intricate patterns in data to predict loan behavior.
- Explainablilty and Causality: For learning the “why” behind the data.
At Menhir we preach by example. And example means economic impact:
Case Study #1: How Algorithmic DCA Allocation Boosted Loan Collections at a Tier-1 Servicer
Overview Banks and Distressed Funds often outsource loan workouts to Servicers, leading to a diluted responsibility across multiple outsourcings and loan managers. The result? Inefficient value chains with middling results, where collection attribution remains a complex task.
The Challenge Menhir aimed to anticipate which loans would yield short-term payments and design an efficient asset management strategy based on algorithmic outputs. With the client holding a +4bn€ NPA portfolio, including both NPL and REO, accuracy was paramount.
The Approach Menhir implemented:
- Payment Anticipation Model: A machine learning model that identifies paying customers before management intervention.
- Allocation Algorithm: Designed to increase management intensity with the same Full-Time Equivalent (FTE) while slashing servicing costs.
The assets were segmented into two categories:
- Assets with Payment Anticipation: for big tickets.
- Assets with Payment Anticipation & Allocation Algorithm: for granular management.
Results over a portfolio of +800M€
The application of AI techniques led to:
- A +50% increase in monthly collection rate.
- A significant reduction in collection cost by 41%.
Conclusion
AI is a beacon of hope for financial institutions grappling with NPL complexities. Menhir’s experience stands as a testament to AI’s transformative power, showcasing its potential in streamlining NPL management, maximizing efficiency, and delivering outstanding results.
Get in touch with our Financial Services practice at fs@menhir.ai
