IFRS 9 was born from the ashes of the 2008 global financial crisis, a time when accounting standards notoriously allowed banks to recognise credit losses only after "the horse had bolted". While the standard has been widely adopted since its inception eight years ago, the maturity of these models remains inconsistent. Many organisations have implemented the standard in form but not in substance, resulting in immature frameworks and significant audit vulnerability. Because IFRS 9 is a set of globally recognised principles that materially impact tier one capital and investor confidence, the gap between compliance and capability remains a significant concern for the industry.
The evolution of 'expected loss'
At its core, the expected credit loss (ECL) framework answers a single question: “How much money do we expect to lose from customers who might not repay?” Instead of waiting for a contractual default, IFRS 9 requires lenders to estimate losses now for events that could happen in the future. To calculate this, organisations combine three key concepts: probability of default (PD), loss given default (LGD) and exposure at default (EAD).
ECL = PD + LGD + EAD
While these components are well-understood, the challenge lies in the interdependency and volatility of these variables under shifting macroeconomic conditions. Moving from recognising losses after the fact to proactive, data-driven forecasting allows businesses to manage credit risk far more effectively. This shift requires a move away from static reporting towards models that factor in economic conditions in real-time.
Why 'good enough' is no longer enough
Many lenders believe IFRS 9 is a "done deal", but the landscape has shifted as the framework enters a more mature phase. Auditor expectations have evolved past the initial implementation learning curve and are now focused on the technical rigour of forward-looking information (FLI) and the justification of manual overlays.
Simultaneously, many models that were built years ago are suffering from 'drift', where the relationship between macroeconomic indicators and default behaviour has fundamentally changed post-pandemic. These poorly calibrated models often lead to over-provisioning, which does not just erode profit; it compromises tier one capital and increases the cost of funding by creating an unnecessary 'risk premium' for wary investors.
Unique challenges for mid-market lenders
Retail and non-bank lenders face uniquely difficult challenges because the nature of their portfolio makes accurate forecasting complex. Unlike traditional banks, they often serve thin-file customers where risk signals are weaker and less stable. At the same time, customer behaviour can shift quickly, with higher volatility in repayment patterns and cure rates. This makes it harder to reliably estimate expected losses using standard approaches.
Compounding this, many organisations operate with limited internal modelling capability, relying on legacy systems or complex spreadsheets that are difficult to maintain, validate and explain. What looks 'good enough' in steady periods can quickly unravel under scrutiny, leading to unexpected provisions and audit findings.
The five most common IFRS 9 failures
Maintaining a robust framework is often undermined by consistent technical and operational failures. Model drift and failed back-tests are leading issues where models are not recalibrated to reflect current behaviour, resulting in outputs that no longer align with reality.
This is frequently compounded by staging and SICR misalignment, where poorly calibrated rules push too many accounts into stage two prematurely, unnecessarily inflating provisions.
Without data-supported forward-looking adjustments explainable within specific macroeconomic scenarios, provisions can appear arbitrary. These gaps are often exacerbated by incomplete documentation and manual, spreadsheet-driven workflows that increase the likelihood of human error and make results difficult to reproduce, raising significant concerns for audit and finance teams alike.
Excel is a liability, not a model
The core argument for a stronger framework is technological. In a modern risk environment, Excel is increasingly viewed as a liability rather than a reliable modelling engine. Moving towards automated data pipelines and code-based modelling in Python or R changes the conversation entirely by providing reproducibility by design. In these environments, every change to an assumption is version-controlled, logged and timestamped, making the framework auditable.
This shift also improves operational efficiency, as scripted pipelines can reduce the 'quarter-end crunch' from days of manual data entry to hours of automated processing. Furthermore, once models sit in code rather than fragile workbooks, running on-demand sensitivity analysis becomes trivial, allowing for genuine forward-looking risk management rather than reactive reporting.
What good looks like
A well-functioning IFRS 9 framework is controlled, transparent and repeatable. It requires auditor-aligned models supported by clear documentation and strong governance, where every assumption is traceable. Strengthening models and automating processes early allows organisations to move from reactive fixes to controlled, predictable outcomes. For mid-market lenders, this is where a practical, analytics-led partner like Principa makes a meaningful difference, helping to stabilise provisions, reduce audit friction and empower business decisions you can trust.
Share
