Chain Forward Risk Viability Assessment — V3.0 Strategic Cut-Off & Viability Path

Model Iterations & Validation: Documenting the Shift from Pricing to Hard Filtering

📌 Project Overview

This repository documents the full analytical journey—from the initial V1.0 feasibility study to the validated V3.0 model deployment strategy. The core objective remains achieving portfolio viability, but the strategic path has evolved from marginal pricing adjustments to implementing a definitive, high-confidence strategic cut-off.

It includes all core project artifacts, model code, and documentation across iterations.

---

Folder Structure

Chain_Forward_Risk_Assessment/
├── 1_Presentation/
│ ├── Achieving Profitability, Risk Model V3.0 Findings & Strategic Cut-Off Recommendation.pptx
│ ├── Chain Forward Profitability Analysis, Strategic Path to Viability.pbix
│ └── Chain Forward Risk Assessment & Profitability Study.pptx
│
├── 2_Code_and_Data/
│ ├── data/
│ │ ├── Loan_Snapshot_Dataset.xlsx
│ │ └── simulated_msme_data.csv
│ │
│ ├── outputs/
│ │ ├── chart_feature_importance.csv
│ │ ├── chart_feature_importance_v3.csv
│ │ ├── chart_segment_performance.csv
│ │ ├── chart_segment_performance_v3.csv
│ │ ├── combined_features_v3.csv
│ │ ├── combined_risk_dataset_final_v4.csv
│ │ ├── sample_combined_v4.csv
│ │ ├── scenario_analysis_results.csv
│ │ ├── scenario_analysis_results_v3.csv
│ │ ├── segmentation_summary.csv
│ │ ├── segmentation_summary_v3.csv
│ │ └── models/
│ │ ├── risk_model_v2.pkl
│ │ └── scaler_v2.pkl
│ │
│ └── src/
│ ├── chain_forward_risk_model.py
│ ├── data_combination_pipeline.py
│ ├── data_combination_pipeline_v2.py
│ └── risk_modeling_pipeline_v2.0.py
│
├── 3_Documentation/
│ ├── Chain Forward Profitability Analysis, Strategic Path to Viability.pdf
│ ├── Model Governance and Strategic Roadmap.pdf
│ └── Risk Management & Monitoring Framework (V3.0 - Post-Deployment).pdf
│
├── 4_Monitoring_Dashboard/
│ └── governance_dashboard.html
│
├── Readme.md
└── run_full_pipeline.ps1

---

🎯 Core Finding (V3.0 Final)

The V3.0 model provides a high-confidence solution:

A Hard Cut-Off (Filtering) of the most loss-concentrated segment (Segment 0: CRITICAL RISK) is required to reduce the portfolio-wide default rate below the 3.75% break-even point and achieve immediate positive NPV.

---

📊 Key Financial Insights (Baseline)

Metric	Result	Implication
Net Present Value (NPV)	−$6,745,425 (3-year horizon)	Product is not financially viable in current form
Break-Even Default Rate	3.75%	Portfolio must stay below this risk level
Current Expected Default Rate	6.00%	Product operates 2.25 percentage points above sustainability

---

🔍 Analytical Findings: Evolution from V1.0 to V3.0

---

V1.0 Findings (Initial Hypothesis & Pricing Strategy)

The initial modeling focused heavily on non-traditional, behavioral variables, finding the model's rejection power was weak (AUC < 0.6). This led to a strategy centered on marginal pricing adjustments.

🧭 V1.0 Methodology

V1.0 followed a structured four-pillar approach:

• Portfolio Viability Modeling
• Behavioral Risk Analytics
• K-Means Segmentation
• Logistic Regression for Default Prediction

🔑 V1.0 Key Analytical Findings

Metric	V1.0 Result	Strategic Recommendation
Dominant Risk Driver	Cashflow Volatility Ratio (+0.3347)	Implement risk-based pricing premiums
Model Limit	ROC AUC = 0.5962 (Weak)	Pricing adjustments needed before stricter modeling
V1.0 Segmentation	Segment 1 (6.66% DR, 48.5% share)	Aggressive pricing required

Loss concentration: Nearly half the borrowers belonged to a high-risk segment above the break-even threshold.

---

V3.0 Findings (Validated Strategy & Hard Filtering)

Refined feature engineering and model tuning yielded a high-confidence predictive model (AUC 0.8820), validating traditional credit signals and enabling a decisive cut-off strategy.

Metric	V3.0 Result	Strategic Implication
Model Performance	ROC AUC = 0.8820 (Excellent)	High-confidence segmentation and rejection rules
Top Risk Drivers	Max Days in Arrears (+1.94), Max Utilization (+1.23)	Traditional credit signals dominate
Loss Concentration	Segment 0: 86.1% DR, 40% volume	Segment 0 must be eliminated

V3.0 Segmentation Summary

Segment	Risk Classification	Default Rate	Portfolio Share
Segment 0	CRITICAL RISK	86.1%	40%
Segment 2	High Risk	61.6%	20%
Segment 1	Base Risk	35.8%	40%

---

🎯 Strategic Recommendations (V3.0 Path to Viability)

The V3.0 strategy is a 3-part plan to immediately reduce portfolio risk and then optimize pricing and retention.

---

📌 Phase 1 — Immediate Strategic Filtering (The Primary Fix)

Objective: Achieve positive NPV by forcing the portfolio DR below 3.75%.

Action	Target Segment	Strategic Action
Filter (Hard Cut-Off)	Segment 0	Auto-decline based on arrears/utilization profile
Price	Segment 2	Apply aggressive risk-based pricing
Retain	Segment 1	Preferential pricing, core segment

Financial Impact Scenario

Scenario	Hypothesis	Post-Filter DR	3-Year NPV
Optimistic	Filter Segment 0 (40% volume)	4.5%	+$750,000
Base Case	No filtering	6.0%	−$6,745,425

---

📌 Phase 2 & 3 — Deployment and Future Modeling

• Deploy the V3.0 model and Segment 0 rejection rule into the lending platform
• Setup Early Warning Indicators
• Future exploration: XGBoost / LightGBM

---

🔄 Automated Execution Pipeline

The repository includes a PowerShell script:

run_full_pipeline.ps1

Runs the entire workflow end-to-end.

▶️ How to Run

.\run_full_pipeline.ps1

---

📂 Pipeline Flow

run_full_pipeline.ps1
│
├── STEP 1 → data_combination_pipeline.py
│       • Load data
│       • Clean & engineer features
│       • Export initial combined dataset
│
├── STEP 2 → data_combination_pipeline_v2.0.py
│       • Refine feature engineering
│       • Apply V3.0 target definition
│       • Export final combined dataset for modeling
│
├── STEP 3 → chain_forward_risk_model.py
│       • K-means segmentation
│       • Logistic regression modeling
│       • Profitability & NPV analysis
│       • Scenario stress-testing
│       • Output generation
│
└── STEP 4 → risk_modeling_pipeline_v2.0.py
        • Model validation and fine-tuning
        • Scenario and stress-test replication
        • Final output generation for monitoring

---

📊 Monitoring Dashboard (Post-Deployment)

File: 4_Monitoring_Dashboard/governance_dashboard.html

To view in browser (Python HTTP server method):

1. Serve the folder from terminal:

cd 4_Monitoring_Dashboard
python -m http.server 8000

2. Open your browser and navigate to:

http://localhost:8000/governance_dashboard.html

⚠️ Using forward slashes / in the URL is essential to avoid 404 errors from the server.

▶️ How to Use

1. Open the link in a web browser.
2. Navigate through tabs to review:
   • Portfolio segmentation and risk exposure
   • Feature importance and model explainability
   • Scenario stress-testing outcomes
3. Use this dashboard for executive reporting and governance purposes.

This allows non-technical stakeholders to monitor portfolio health and model performance without needing to run scripts.

---

📁 Repository Structure

Folder	Description
1_Presentation/	Final PowerPoint presentation and Power BI visuals
2_Code_and_Data/src/	Core Python scripts for modeling and scenario analysis
2_Code_and_Data/outputs/	CSV outputs for segmentation, profitability, and feature weights
3_Documentation/	Supporting notes, assumptions, frameworks

---

🛠 Technical Stack

Tool	Purpose
Python	Core analytics & modeling
Pandas / NumPy	Data engineering & financial calculations
Scikit-learn	Clustering + Logistic Regression modeling
Power BI	Executive-ready dashboard and visualization