Commercial Exposure at Default & CCF Project

This repository is the EAD and Credit Conversion Factor layer in the public commercial credit-risk stack. It uses synthetic facility data, utilisation assumptions, and product-level CCF logic to estimate funded and unfunded exposure under lending workflows that are relevant to both bank-style risk frameworks and practical lending decisioning. The main outputs feed downstream expected loss, stress testing, pricing, and capital analysis.

What this repo is

This project demonstrates how a commercial lending portfolio can be translated into Exposure at Default measures using transparent, recruiter-friendly logic. It is built as a portfolio project rather than a production bank engine, so the assumptions are explainable and the data is synthetic where internal usage and drawdown history would normally be required.

Where it sits in the stack

Upstream inputs:

• facility and limit inputs staged under data/
• product utilisation and CCF assumptions maintained in-repo

Downstream consumers:

• expected-loss-engine-commercial
• stress-testing-commercial
• RAROC-pricing-and-return-hurdle
• RWA-capital-commercial

How this is used in practice

This project can be applied in:

Bank / Institutional context

• EAD estimation for expected loss, stress testing, and capital-style frameworks
• Portfolio exposure measurement across funded and unfunded facilities
• Utilisation and limit analysis for structured risk review

Non-bank / Fintech context

• Exposure measurement for pricing and approval strategy on revolving or contingent products
• Limit-usage assumptions for portfolio performance and risk-adjusted decisioning
• Early portfolio monitoring of drawn versus undrawn exposure risk

Example input files (already in the repo)

• data/raw/exposure_master.csv: demo facility-level limits and drawn balances (funded + unfunded)
• data/raw/demo_portfolio.csv: lightweight portfolio extract used by the demo pipeline
• data/manual/ccf_rules.csv: product-level CCF rules and utilisation assumptions

Example output files (already in the repo)

• outputs/tables/ead_by_facility.csv: facility-level EAD view used downstream
• outputs/tables/ccf_by_product.csv: product-level CCF summary (sanity-check friendly)
• outputs/tables/utilisation_uplift_tables.csv: utilisation and uplift curves used by the EAD build
• outputs/tables/pipeline_validation_report.csv: pass/fail checks for required fields and totals
• outputs/reports/pipeline_summary.md: short run summary and file index
• outputs/samples/demo_input.csv: small sample input for quick reviewer inspection

Example business use case

A credit portfolio team needs a consistent “exposure at default” view before building expected loss, stress tests, pricing packs, or capital summaries. This repo turns limits + drawn balances into a facility-level EAD dataset with a clear, reviewable CCF contract.

How these outputs feed downstream repos

• expected-loss-engine-commercial: uses outputs/tables/ead_by_facility.csv as the exposure leg of EL.
• stress-testing-commercial: uses outputs/tables/ead_by_facility.csv as the base EAD input before applying scenario uplifts.
• RAROC-pricing-and-return-hurdle: uses EAD-derived exposure to translate EL and stress into cost-of-risk and hurdle pricing.
• RWA-capital-commercial: uses outputs/tables/ead_by_facility.csv as the exposure input for RWA and capital summaries.

Key inputs

• facility and borrower portfolio data
• product limit, drawn-balance, and limit-utilisation information
• CCF and utilisation assumptions where detailed behavioural history is unavailable

Key outputs

• outputs/tables/ead_by_facility.csv
• outputs/tables/ccf_by_product.csv
• outputs/tables/utilisation_uplift_tables.csv
• outputs/tables/ead_validation_report.csv
• outputs/tables/pipeline_validation_report.csv

Repo structure

• data/: raw, interim, processed, and external demo inputs
• src/: reusable EAD and CCF pipeline logic
• scripts/: wrapper scripts for running the pipeline
• docs/: methodology, assumptions, data dictionary, and validation notes
• notebooks/: reviewer-facing walkthrough notebooks
• outputs/: exported tables, reports, and sample artifacts
• tests/: validation and regression checks

How to run

Quick start:

pip install -r requirements.txt
python -m src.run_pipeline

After the run, start with:

• outputs/reports/pipeline_summary.md
• outputs/tables/ead_by_facility.csv
• outputs/tables/pipeline_validation_report.csv

Run validation tests:

python -m pytest

Alternative (wrapper script):

python scripts/run_demo_pipeline.py

Testing and validation

• tests/test_demo_pipeline.py runs a minimal demo pipeline and asserts the expected output files are written.
• outputs/tables/pipeline_validation_report.csv captures the same checks in a reviewer-friendly table.

Limitations / Demo-Only Note

• All portfolio inputs are synthetic and are included for demonstration only.
• CCF and utilisation behaviour is modelled with simplified, transparent assumptions rather than internal behavioural datasets.
• The repo is intended for portfolio presentation and workflow demonstration, not for production limit management or regulatory reporting.