Skip to content
Site search

Find services, work and insights

PD, LGD, EAD and CCF

Credit risk modelling

Develop, recalibrate and document transparent credit risk parameter models for decision-making, provisioning and capital assessment.

Regulatory context

What makes the work necessary

Credit risk parameters need to reflect portfolio behaviour, withstand independent review and remain reproducible after handover. The engagement joins statistical development with implementation controls and documentation.

Common challenges

Signals that the current approach needs attention

Parameter estimates no longer reflect observed portfolio behaviour

Segmentation and default definitions are not consistently implemented

Model code cannot be reproduced or independently reviewed

Performance, calibration and stability evidence is incomplete

Intended outcomes

Capabilities the engagement is designed to create

  • Transparent parameter methodology linked to the business question
  • Reproducible estimation and monitoring code
  • Performance evidence covering discrimination, calibration and stability
  • Audit-ready methodology and implementation documentation
Deliverables

Tangible outputs for analysis and handover

Data and definition assessment
Segmentation analysis
PD, LGD, EAD or CCF model implementation
Calibration and backtesting pack
Methodology and limitations report
Methodology

Structured from framing to transfer

Each stage leaves an output that can be reviewed before the next stage begins.

01 Frame Clarify the question, users, outputs and governance requirements.
02 Assess Review data, methods, tools, controls and model risk.
03 Model Prepare data, test assumptions and build transparent analysis.
04 Validate Challenge performance, stability, implementation and limitations.
05 Automate Add repeatable code, controls, reconciliation and logging.
06 Transfer Document, train, hand over and define monitoring.
Models and methods

Methods selected around the analytical purpose

  • Probability of Default
  • Loss Given Default
  • Exposure at Default
  • Credit Conversion Factor
  • Logistic regression
  • Calibration
  • Discrimination
  • Stability analysis
  • Backtesting
Technology

The tool follows the control environment

  • Python
  • R
  • SQL
  • SAS
Frameworks
  • IFRS 9
  • Basel frameworks
  • ICAAP
Data requirements

Evidence needed to support the scope

  • Documented portfolio, facility and default data with lineage
  • Sufficient outcome history for the selected estimation approach
  • Agreed definitions, exclusions and reference dates
Governance

Controls that preserve accountability

  • Model purpose, scope and material limitations are recorded
  • Development and validation evidence is separated where required
  • Code, assumptions and data transformations are version-controlled
Related experience

Source-verified implementation examples

Frequently asked

Before scoping the work

Can an existing model be recalibrated without replacing it?

Yes. The first step is to assess conceptual soundness, data changes and performance evidence, then determine whether recalibration is sufficient or redevelopment is more appropriate.

Do you support independent model validation?

Yes. Validation scope can cover methodology, data, implementation, discrimination, calibration, stability, backtesting and documentation.

Next step

Discuss a credit risk modelling requirement.

A short brief is enough to establish whether this is a fit. Describe the situation in general terms only. Do not send client data or model files.

Privacy controls