IFRS 9 and CECL

IFRS 9 (Financial Instruments), effective January 1, 2018 for most entities, replaced IAS 39 and introduced the Expected Credit Loss (ECL) model for impairment, replacing the incurred loss model. Under IFRS 9, financial assets are classified into three stages: Stage 1 (performing assets) requires 12-month ECL; Stage 2 (significant increase in credit risk since origination, SICR) requires lifetime ECL; Stage 3 (credit-impaired) requires lifetime ECL with interest calculated on net carrying amount. The US equivalent, CECL (Current Expected Credit Loss) under ASC 326, became effective for SEC filers in 2020. While IFRS 9 uses a three-stage approach, CECL requires lifetime ECL measurement for all financial assets from Day 1 — a more conservative approach. Both standards require forward-looking macroeconomic scenarios to be incorporated into ECL calculations, require the use of all available "reasonable and supportable" information, and require extensive disclosure of model assumptions, sensitivities, and methodology.

IFRS 9 and CECL are fundamentally data and modeling problems. ECL models require: probability of default (PD) estimates at origination and at each reporting date; loss given default (LGD) estimates incorporating collateral values and recovery rates; exposure at default (EAD) incorporating undrawn commitments and credit conversion factors; and macroeconomic variable sensitivities that link ECL outputs to economic scenarios (GDP growth, unemployment, property prices). For large banks with diverse portfolios, ECL model suites encompass hundreds of models by product type, geography, and vintage. These models require vintage-level historical loan performance data going back multiple economic cycles (pre-2008 data is critical for through-the-cycle calibration). SICR assessment under IFRS 9 requires automated comparison of PD at origination against current PD across every performing loan in the portfolio — a computation that must complete within the monthly close cycle for entities with millions of loan records.

IFRS 9 and CECL interact with stress testing frameworks in ways that create modeling system architecture constraints. CCAR/DFAST stress testing uses stressed PD/LGD/EAD inputs for regulatory scenarios, while IFRS 9/CECL uses probability-weighted scenarios for accounting provisioning — the underlying models may be shared or separate, but results must be reconcilable. The COVID-19 pandemic revealed a critical modeling challenge: both standards depend on historical data patterns, and a novel shock with no historical precedent required management overlays and expert judgment adjustments that models could not provide. Management overlay governance — the process for documenting, approving, and disclosing expert adjustments to model outputs — became a significant audit focus area. For IFRS 9, the classification and measurement rules also drive technology requirements: the SPPI (solely payments of principal and interest) test for business model assessment of financial assets requires systematic analysis of contractual cash flow characteristics at origination.

We build ECL calculation platforms with model-agnostic execution engines that support PD/LGD/EAD model plug-ins by portfolio segment, scenario weighting frameworks for probability-weighted ECL, and staging logic implementing IFRS 9 SICR thresholds with configurable backstop criteria. Our platforms produce loan-level ECL output with full calculation audit trails, supporting both IFRS 9 disclosure requirements and CCAR/DFAST stress testing reconciliation. Management overlay workflows are implemented with structured justification capture, approval routing, and quantitative impact attribution to ensure overlay governance meets auditor and regulatory scrutiny standards.

Compliance-Native Architecture Guide

Design principles and a structured checklist for building software that is compliant by default — not compliant by retrofit. Covers data architecture, access controls, audit trails, and vendor due diligence.