# Expected Credit Loss Methodology: [Entity or portfolio]

> Methodology document for measuring loss allowances on loans, debt
> securities, loan commitments and financial guarantees: scope and
> framework, governance, segmentation, definition of default, staging and
> significant increase in credit risk (SICR), the probability of default
> (PD) x loss given default (LGD) x exposure at default (EAD) x discount
> factor calculation, forward-looking macroeconomic scenarios and their
> weights, component models, individually assessed exposures, management
> overlays, a US current expected credit loss (CECL) variant, validation,
> disclosure and a worked example. Written for banks and lenders that
> report under IFRS 9 Financial Instruments section 5.5 or under US GAAP
> FASB ASC Topic 326; both are cited by name and paragraph or topic
> number only, and supervisory expectations come from EBA/GL/2017/06.
> Replace every `[bracketed]` value and every row marked *Example*. This
> is a document format, not accounting, financial, audit or regulatory
> advice. Requirements are referenced by number: obtain the standard for
> the normative text.

<!--
Guidance: How to use this template
- This is the bank's own methodology: it records the choices the bank
  made, the evidence for them and who approved them. Auditors and
  supervisors test the allowance against this document, so every
  judgement in the calculation should be traceable to a section here.
- Read IFRS 9 section 5.5 (or ASC 326-20) yourself before you complete
  sections 4 to 8. This template deliberately does not restate the
  standards; it cites paragraph numbers so you can find the requirement.
- Fill sections 1 to 3 once, then sections 4 to 8 per portfolio
  segment. Keep thresholds and parameters in the parameter register
  (section 8.4) so a change is versioned and approved, not edited in code.
- Worked example: the fictional Fernhollow Bank (mid-sized EU credit
  institution, IFRS reporter, EUR, financial year = calendar year). The
  stage 3 example is the Quillmoor Logistics GmbH case from Problem Loan
  Workout Plan (`finance/problem-loan-workout-plan`). All figures are
  invented and every derived number has been recomputed.
- Related documents: Model Development Document
  (`finance/model-development-document`), Model Risk Management Policy
  (`finance/model-risk-management-policy`), Model Validation Report
  (`ai-data/model-validation-report`), Stress Testing Programme
  (`finance/stress-testing-programme`), Collateral Valuation Review
  (`finance/collateral-valuation-review`), Annual Credit Review
  (`finance/annual-credit-review`), Accounting Policy Manual
  (`finance/accounting-policy-manual`) and Technical Accounting Memo
  (`finance/technical-accounting-memo`).
-->

## Document control

<!-- Guidance: The methodology is a controlled document. Changes to staging thresholds, scenario weights or model components need the same approval as the original. Pitfall: changing a parameter in the production engine without a methodology version. -->

| Field | Value |
|---|---|
| Document ID | [ECL-METH-NNN] *Example:* ECL-METH-003 |
| Entity and reporting framework | [Legal entity, IFRS / US GAAP] *Example:* Fernhollow Bank, consolidated, IFRS as adopted by the EU |
| Portfolios in scope | [List] *Example:* Corporate and SME loans, residential mortgages, consumer loans, debt securities at amortised cost and FVOCI, loan commitments and guarantees |
| Version | [3.0] |
| Status | [Draft / In review / Approved / Superseded] |
| Owner | [@head-of-credit-risk-modelling] |
| Approvers/reviewers | [@cfo], [@cro], [@ecl-committee], [@model-validation], [@audit-committee] (informed) |
| Effective from reporting date | YYYY-MM-DD *Example:* 2026-12-31 |
| Last updated | YYYY-MM-DD *Example:* 2026-10-20 |
| Classification | [Internal] |
| Currency and units | [EUR million for portfolio figures, EUR thousand for loan-level examples] |
| Next scheduled review | YYYY-MM-DD *Example:* 2027-10-31 |

### Approvals

| Role | Name | Decision | Date |
|---|---|---|---|
| Methodology owner | [@name] | [Proposed] | YYYY-MM-DD |
| Independent model validation | [@name] | [Validated / Validated with findings] | YYYY-MM-DD |
| ECL committee (CFO and CRO co-chairs) | [@name] | [Approved] | YYYY-MM-DD |
| Audit committee | [@name] | [Noted] | YYYY-MM-DD |

### Revision history

| Version | Date | Author | Change |
|---|---|---|---|
| 2.0 | YYYY-MM-DD | [@author] | Annual review |
| 3.0 | YYYY-MM-DD | [@author] | [Change summary] *Example:* New downside scenario weight, SICR absolute threshold added |

## Contents

- [1. Purpose, scope and framework](#1-purpose-scope-and-framework)
- [2. Governance and roles](#2-governance-and-roles)
- [3. Portfolio segmentation](#3-portfolio-segmentation)
- [4. Definition of default and credit-impaired](#4-definition-of-default-and-credit-impaired)
- [5. Staging and significant increase in credit risk](#5-staging-and-significant-increase-in-credit-risk)
- [6. Measurement](#6-measurement)
- [7. Forward-looking information and scenarios](#7-forward-looking-information-and-scenarios)
- [8. Component models and parameters](#8-component-models-and-parameters)
- [9. Individually assessed exposures](#9-individually-assessed-exposures)
- [10. Management overlays](#10-management-overlays)
- [11. US GAAP CECL variant](#11-us-gaap-cecl-variant)
- [12. Validation, back-testing and monitoring](#12-validation-back-testing-and-monitoring)
- [13. Reporting and disclosure](#13-reporting-and-disclosure)
- [14. Glossary and references](#14-glossary-and-references)
- [Standard and licence note](#standard-and-licence-note)

## 1. Purpose, scope and framework

<!-- Guidance: Define which instruments are in scope and which framework applies to each entity in the group. Pitfall: forgetting off-balance-sheet commitments and guarantees, or intercompany loans in solo accounts. -->

State what the methodology covers, which accounting framework applies,
and how this document relates to the bank's accounting policies, credit
risk policies and regulatory reporting. Where the group has entities
reporting under different frameworks, describe both and reconcile them.

| Instrument | IFRS 9 reference | In scope *(Example)* |
|---|---|---|
| Loans and advances at amortised cost | Section 5.5 | Yes |
| Debt securities at FVOCI | Section 5.5 (allowance recognised in OCI) | Yes |
| Loan commitments not at FVTPL | Paragraph 5.5.20 for revolving facilities | Yes |
| Financial guarantee contracts | Section 5.5 | Yes |
| Trade and lease receivables | Paragraph 5.5.15 (simplified approach) | Yes, simplified approach for trade receivables of the leasing subsidiary |
| Purchased or originated credit-impaired (POCI) | Paragraph 5.5.13 | Yes, acquired NPL portfolio |
| Financial assets at FVTPL | Outside the impairment requirements | No |

| Dimension | IFRS 9 (section 5.5) | US GAAP CECL (ASC 326-20) |
|---|---|---|
| Allowance horizon | 12-month expected losses for stage 1, lifetime for stages 2 and 3 | Lifetime from initial recognition for all assets in scope |
| Staging | Three-stage practice built around SICR | No staging |
| Forward-looking information | Reasonable and supportable information, multiple scenarios where loss is non-linear | Reasonable and supportable forecast period, then reversion to historical loss information |
| Discounting | At the effective interest rate when a cash-flow approach is used | Required only for discounted cash flow methods |
| Off-balance-sheet exposures | Loan commitments and guarantees | Unconditionally cancellable commitments excluded |
| Supervisory guidance used here | EBA/GL/2017/06 | Interagency Policy Statement on Allowances for Credit Losses (SR 20-12, revised 2023-04-21) |

## 2. Governance and roles

<!-- Guidance: Name who owns each component, who validates, who approves overlays and who signs the final allowance. Pitfall: the same team builds, runs and approves the models with no independent challenge. -->

EBA/GL/2017/06 expects a credit risk management framework that supports
the ECL estimate, with clear responsibilities, documented methodologies,
independent validation and senior management oversight. Describe the
process from data cut to booked allowance and the controls at each step.

| Role | Responsibility | *Example* holder |
|---|---|---|
| Methodology owner | Maintains this document, proposes changes | Head of Credit Risk Modelling |
| Model owners | PD, LGD, EAD and scenario models | Modelling team leads |
| Economics function | Macroeconomic scenarios and weights | Chief Economist |
| Credit risk officers | Individual assessments, watch list, overlays evidence | Credit Risk |
| Finance | Runs the engine, posts the allowance, prepares disclosures | Financial Reporting |
| Model validation | Independent validation and annual review | Model Risk |
| ECL committee | Approves scenarios, weights, overlays and the final allowance each quarter | CFO and CRO co-chairs |
| Internal audit | Periodic audit of the process | Internal Audit |

```mermaid
flowchart LR
  A["Data cut at reporting date"] --> B["Staging engine"]
  C["Scenarios and weights approved"] --> D["PD, LGD, EAD per scenario"]
  B --> E["ECL engine"]
  D --> E
  F["Individual assessments stage 3"] --> G["Model output plus individual"]
  E --> G
  G --> H["Overlays proposed with evidence"]
  H --> I["ECL committee approval"]
  I --> J["Allowance booked and disclosed"]
```

## 3. Portfolio segmentation

<!-- Guidance: Group exposures with shared credit risk characteristics, and keep segments stable over time. Too many segments leave thin data; too few hide different risk drivers. Pitfall: segmenting by product system rather than by risk driver. -->

Define segments by shared risk characteristics such as borrower type,
product, collateral, industry, geography and rating method. For each
segment, record the model set used and the share of the portfolio.

| Segment ID | Segment | Gross carrying amount (EUR million) *(Example)* | Models | Collateral |
|---|---|---|---|---|
| SEG-01 | Large corporate | 2,450 | Rating-based PD, workout LGD | Mixed |
| SEG-02 | SME | 1,620 | Scorecard PD, collateral LGD | Mostly real estate |
| SEG-03 | Commercial real estate | 1,180 | Rating-based PD, collateral LGD | Real estate |
| SEG-04 | Residential mortgages | 3,410 | Behavioural PD, collateral LGD with house price index | Residential property |
| SEG-05 | Consumer and cards | 900 | Behavioural PD, cure-rate LGD | Unsecured |
| **Total loans** | | **9,560** | | |

The total equals the stage table in section 6.4 (8,200 + 1,100 + 260 =
9,560).

Review the segmentation each year. A segment whose default experience
diverges from its neighbours, or whose collateral or borrower mix has
changed, may need to be split; a segment with too few defaults to estimate
parameters reliably may need to be merged or supported by external data.
Document every change, run the old and new segmentation in parallel for at
least one quarter, and disclose the effect on the allowance if it is
material.

## 4. Definition of default and credit-impaired

<!-- Guidance: Use one definition of default for accounting, risk management and regulatory purposes unless you can justify a difference. Pitfall: a different default definition in the PD model data from the one used in staging. -->

IFRS 9 does not define default; it requires an entity to use a definition
consistent with its credit risk management and sets a rebuttable
presumption at more than 90 days past due (paragraph B5.5.37). Most EU
banks align the accounting definition with the prudential definition in
CRR Art. 178 and EBA/GL/2016/07, so that stage 3 and default broadly
coincide. Record your definition and every difference from the prudential
one.

| Element | Fernhollow policy *(Example)* | Basis |
|---|---|---|
| Past due criterion | More than 90 days past due on a material amount | CRR Art. 178(1)(b); Delegated Regulation (EU) 2018/171 thresholds |
| Unlikeliness to pay | Non-accrual, specific credit risk adjustment, sale at a material credit loss, distressed restructuring with diminished obligation above 1%, insolvency filing | CRR Art. 178(3); EBA/GL/2016/07 |
| Contagion | All exposures of a defaulted corporate obligor; retail by facility with pulling effect rule | EBA/GL/2016/07 |
| Exit from default | Minimum 3 months probation, 12 months after distressed restructuring | EBA/GL/2016/07 (as amended by EBA/GL/2026/05, applying from 2026-10-19) |
| Credit-impaired (stage 3) | Equal to default | IFRS 9 Appendix A definition; alignment documented |
| Write-off | When there is no reasonable expectation of recovery, for example after enforcement ends | IFRS 9 paragraph 5.4.4 |

Interest revenue on stage 3 assets is calculated on the amortised cost net
of the allowance (IFRS 9 paragraph 5.4.1). Make sure the engine switches
basis when an exposure moves into or out of stage 3.

## 5. Staging and significant increase in credit risk

<!-- Guidance: SICR compares the risk of default now with the risk of default at initial recognition, over the remaining life. Combine a quantitative test with qualitative indicators and the 30-days-past-due backstop. Pitfall: an absolute PD threshold only, which keeps high-risk originations in stage 1 forever. -->

IFRS 9 paragraphs 5.5.3, 5.5.5 and 5.5.9 drive the stage split; paragraph
5.5.10 permits a low credit risk simplification and paragraph 5.5.11 sets
a rebuttable 30-days-past-due presumption. Describe each criterion, its
threshold and the evidence that calibrates it. EBA/GL/2017/06 discourages
wide use of the low credit risk simplification by banks for loans.

### 5.1 Criteria

| Criterion ID | Test | Threshold *(Example)* | Applies to |
|---|---|---|---|
| SICR-Q1 | Relative increase in remaining lifetime PD since origination | Current lifetime PD at least 2.5 times the origination lifetime PD for the same remaining term | All modelled segments |
| SICR-Q2 | Absolute increase in remaining lifetime PD | At least 0.50 percentage points (must also meet SICR-Q1) | All modelled segments |
| SICR-Q3 | Rating downgrade (where lifetime PD is unavailable) | Three grades down on the FH-1 to FH-10 master scale (Credit Rating Rationale, `finance/credit-rating-rationale`) | Large corporate legacy loans |
| SICR-QL1 | Watch list | Placed on watch list or close monitoring (Annual Credit Review, `finance/annual-credit-review`, section 9 categories) | Corporate, SME, CRE |
| SICR-QL2 | Forbearance | Forborne performing flag | All |
| SICR-B1 | Backstop | More than 30 days past due | All |
| SICR-L1 | Low credit risk exemption | Investment-grade debt securities only | Securities portfolio |

### 5.2 Transfers back

| Rule | Fernhollow policy *(Example)* |
|---|---|
| Stage 2 to stage 1 | All SICR triggers cleared at the reporting date; forborne exposures only after the probation period ends |
| Stage 3 to stage 2 | Default exit criteria met (section 4) |
| Quarterly reassessment | Every exposure re-staged at each reporting date |

*Example* SICR test: a corporate loan had a remaining lifetime PD of 3.0%
at origination for its remaining term; the current lifetime PD is 8.1%.
The ratio is 8.1 / 3.0 = 2.7, above 2.5, and the absolute increase is 5.1
percentage points, above 0.50. The loan moves to stage 2.

```mermaid
stateDiagram-v2
  [*] --> Stage1: "Initial recognition"
  Stage1 --> Stage2: "SICR or more than 30 days past due"
  Stage2 --> Stage1: "Triggers cleared, probation met"
  Stage2 --> Stage3: "Default"
  Stage1 --> Stage3: "Default"
  Stage3 --> Stage2: "Exit from default"
  Stage3 --> WrittenOff: "No reasonable expectation of recovery"
```

Test the staging rules every year: how many exposures reach stage 2 before
default (a good rule moves most defaults through stage 2 first), how long
they stay there, and how often exposures move back and forth. Many direct
moves from stage 1 to stage 3 suggest the SICR thresholds are too loose.

## 6. Measurement

<!-- Guidance: Measure an unbiased, probability-weighted amount that reflects the time value of money and reasonable and supportable information (IFRS 9 paragraph 5.5.17). Show the formula, the time grid and the discounting convention. Pitfall: discounting at a market rate instead of the effective interest rate. -->

### 6.1 Formula

For each exposure, scenario $s$ and future period $t$:

$$
\text{ECL}_s = \sum_{t=1}^{T} \text{PD}^{\text{marg}}_{s,t} \times \text{LGD}_{s,t} \times \text{EAD}_{t} \times \frac{1}{(1+\text{EIR})^{t}}
$$

$$
\text{PD}^{\text{marg}}_{s,t} = \text{PD}^{\text{cond}}_{s,t} \times \prod_{k=1}^{t-1}\left(1 - \text{PD}^{\text{cond}}_{s,k}\right)
$$

$$
\text{ECL} = \sum_{s} w_s \times \text{ECL}_s, \qquad \sum_s w_s = 100\%
$$

$T$ is one year for stage 1 (12-month ECL) and the remaining expected life
for stages 2 and 3. The marginal PD is the probability, seen from the
reporting date, of surviving to the start of period $t$ and defaulting
during it; the conditional (hazard) PD is the probability of default in
period $t$ given survival to its start.

### 6.2 Conventions

| Convention | Fernhollow choice *(Example)* |
|---|---|
| Time grid | Annual for corporate, monthly for retail |
| Timing of default | End of period, before the scheduled repayment |
| EAD in period | Balance at the start of the period plus expected drawdowns |
| Discount rate | Original EIR of each exposure (credit-adjusted EIR for POCI) |
| Maximum period | Contractual maturity, or behavioural life for revolving facilities (paragraph 5.5.20) |
| Currency | Calculated in exposure currency, translated at the closing rate |

### 6.3 Worked example (single loan)

*Example* loan: EUR 1,000 thousand, EIR 4.0%, three years remaining,
repaid in three equal annual instalments, so EAD is 1,000.00, 666.67 and
333.33 at the start of years 1, 2 and 3. The loan is in stage 2, so the
lifetime ECL applies.

| Scenario | Year | Conditional PD | Marginal PD | LGD | EAD | Discount factor | ECL |
|---|---|---|---|---|---|---|---|
| Base (weight 50%) | 1 | 2.00% | 2.0000% | 40% | 1,000.00 | 0.9615 | 7.69 |
| Base | 2 | 2.50% | 2.4500% | 40% | 666.67 | 0.9246 | 6.04 |
| Base | 3 | 3.00% | 2.8665% | 40% | 333.33 | 0.8890 | 3.40 |
| **Base lifetime** | | | | | | | **17.13** |
| Upside (weight 20%) | 1 | 1.20% | 1.2000% | 35% | 1,000.00 | 0.9615 | 4.04 |
| Upside | 2 | 1.50% | 1.4820% | 35% | 666.67 | 0.9246 | 3.20 |
| Upside | 3 | 1.80% | 1.7517% | 35% | 333.33 | 0.8890 | 1.82 |
| **Upside lifetime** | | | | | | | **9.05** |
| Downside (weight 30%) | 1 | 4.50% | 4.5000% | 50% | 1,000.00 | 0.9615 | 21.63 |
| Downside | 2 | 5.50% | 5.2525% | 50% | 666.67 | 0.9246 | 16.19 |
| Downside | 3 | 6.50% | 5.8661% | 50% | 333.33 | 0.8890 | 8.69 |
| **Downside lifetime** | | | | | | | **46.51** |

Check one line: base year 2 marginal PD is 2.50% x (1 less 2.00%) =
2.45%; ECL is 0.0245 x 0.40 x 666.67 x 1 / 1.04^2 = 6.04.

Probability-weighted lifetime ECL = 0.50 x 17.13 + 0.20 x 9.05 + 0.30 x
46.51 = 8.57 + 1.81 + 13.95 = 24.33, or 2.43% of the exposure. If the loan
were in stage 1, only year 1 would count: 0.50 x 7.69 + 0.20 x 4.04 + 0.30
x 21.63 = 11.14. Note that the weighted result (24.33) is higher than the
base-scenario result (17.13): losses rise faster in the downside than they
fall in the upside, which is why a single central scenario understates
ECL.

### 6.4 Portfolio result

| Stage | Gross carrying amount (EUR million) *(Example)* | Coverage | Allowance (EUR million) |
|---|---|---|---|
| Stage 1 | 8,200 | 0.25% | 20.5 |
| Stage 2 | 1,100 | 2.80% | 30.8 |
| Stage 3 | 260 | 38.00% | 98.8 |
| **Total** | **9,560** | **1.57%** | **150.1** |

The stage 3 share is 260 / 9,560 = 2.7% of customer loans at amortised
cost. The gross NPL ratio of 2.1% reported in the Recovery Plan
(`finance/recovery-plan`, RI-10) uses the wider EBA denominator, which
also includes loans and advances to banks and balances at central banks;
reconcile the two in the disclosure so that readers do not see two
unexplained figures.

## 7. Forward-looking information and scenarios

<!-- Guidance: Use a small set of internally consistent scenarios that span the range of plausible outcomes, with weights justified by evidence. Use the same scenarios as planning and stress testing where you can, and explain any difference. Pitfall: a severe stress scenario used as an ECL downside, which is not an unbiased estimate. -->

Describe how scenarios are produced, how often, how many, what variables
they contain, how far the explicit forecast runs and how the variables
revert to long-run averages. Then explain how the weights are set and
approved. Link each variable to the PD and LGD models through documented
satellite models.

| Variable (euro area, annual) | Base 2027 | Base 2028 | Upside 2027 | Downside 2027 | Long-run anchor |
|---|---|---|---|---|---|
| Real GDP growth *(Example)* | 1.2% | 1.4% | 2.3% | minus 1.5% | 1.3% |
| Unemployment rate *(Example)* | 6.4% | 6.3% | 5.9% | 8.2% | 6.5% |
| Residential property prices *(Example)* | 2.0% | 2.5% | 4.5% | minus 8.0% | 2.0% |
| Commercial property prices *(Example)* | minus 1.0% | 1.5% | 2.0% | minus 15.0% | 1.0% |
| 3-month interbank rate *(Example)* | 2.0% | 2.0% | 2.3% | 1.2% | 2.0% |
| **Weight** *(Example)* | **50%** | | **20%** | **30%** | |

The *Example* weights sum to 100% (50% + 20% + 30%). Fernhollow sets the
downside weight above the upside weight because the economics function
assesses the distribution of outcomes as skewed to the downside; record the
evidence for that view, such as the distribution of external forecasts. A
change of weight between quarters needs a reason and the effect on the
allowance.

| Sensitivity *(Example)* | Allowance (EUR million) | Change |
|---|---|---|
| Reported (weighted) | 150.1 | |
| 100% base scenario | [Engine output] | [Difference] |
| 100% downside scenario | [Engine output] | [Difference] |
| Downside weight 40%, base 40% | [Engine output] | [Difference] |

## 8. Component models and parameters

<!-- Guidance: Explain how each regulatory or scorecard model is adapted for accounting: point-in-time not through-the-cycle, no downturn or conservatism add-ons, lifetime term structures, forward-looking adjustment. Pitfall: reusing IRB parameters without removing prudential margins. -->

### 8.1 PD

Describe the source model (rating or behavioural scorecard), the
conversion from through-the-cycle to point-in-time PDs, how the term
structure is built (for example migration matrices or survival curves) and
how macroeconomic variables shift it in each scenario. Remove prudential
floors and margins of conservatism required only for capital purposes.

Explain how the PD responds to the economy. A satellite model links the
default rate of each segment to the scenario variables, estimated on a long
enough history to include at least one downturn. Check the sign and size of
each coefficient for economic sense, and document what happens beyond the
explicit forecast horizon, where PDs revert to their long-run average.

### 8.2 LGD

Describe the LGD approach per segment: workout LGD from historical
recoveries, or collateral-based LGD from collateral values, haircuts,
time to recovery and costs, with cure rates for retail. Link collateral
inputs to Collateral Valuation Review (`finance/collateral-valuation-review`)
and to the house and commercial property price paths in each scenario.
Discount recoveries at the EIR, not at a regulatory discount rate.

$$
\text{LGD} = 1 - \frac{\text{PV of expected recoveries net of costs}}{\text{EAD}}
$$

### 8.3 EAD

Describe the repayment profile for amortising loans, prepayment
assumptions, and credit conversion factors for undrawn commitments. For
revolving facilities such as credit cards and overdrafts, explain the
behavioural life used under paragraph 5.5.20 and the evidence for it, for
example the observed time from a credit deterioration signal to default or
limit reduction.

### 8.4 Parameter register

| Parameter ID | Parameter | Value *(Example)* | Evidence | Approved |
|---|---|---|---|---|
| PAR-01 | SICR relative threshold | 2.5x | Back-test 2021 to 2025: 78% of defaults passed through stage 2 at least 6 months before default | ECL committee 2026-09 |
| PAR-02 | SICR absolute threshold | 0.50 percentage points | Removes noise for very low PDs | ECL committee 2026-09 |
| PAR-03 | Scenario weights | 50 / 20 / 30 | Economics function paper Q3 2026 | ECL committee 2026-09 |
| PAR-04 | Revolving behavioural life, cards | 36 months | Observed account behaviour | ECL committee 2026-03 |
| PAR-05 | CRE forced-sale discount in LGD | 20% | Back-tested recoveries | Credit risk 2026-06 |

## 9. Individually assessed exposures

<!-- Guidance: Large stage 3 exposures are usually assessed one by one with scenario-weighted cash flows. Use the same discount rate and recovery assumptions as the workout plan. Pitfall: using only the most likely workout outcome, which ignores re-default and enforcement paths. -->

Define the threshold for individual assessment and the method. For each
exposure, estimate expected cash flows under at least two outcomes
(for example successful restructuring and enforcement), weight them, and
discount at the original EIR. Use the latest collateral valuation and
document the assumptions in the workout plan.

| Item | *Example* (Quillmoor Logistics GmbH, EUR thousand) |
|---|---|
| Threshold for individual assessment | Stage 3 exposures above EUR 1 million per obligor group |
| Gross carrying amount after modification | 18,936.2 (modification loss 1,063.8 recognised under paragraph 5.4.3) |
| Expected present value of recoveries | 15,927.6 (85% restructuring performs, 15% re-default and enforcement) |
| Allowance | 3,008.6 (18,936.2 less 15,927.6) |
| Total credit loss recognised | 4,072.4, or 20.36% of the original 20,000 exposure |
| Source | WO-2026-014, CVR-2026-031 |

## 10. Management overlays

<!-- Guidance: An overlay corrects for a known limitation of the models: a risk the models do not capture, data that are not yet available, or an event after the data cut. Each overlay needs a quantified basis, an owner and a release plan. Pitfall: permanent overlays that have become an unexplained buffer. -->

Overlays, also called post-model adjustments, should be rare, targeted and
temporary. EBA/GL/2017/06 expects them to be well documented and subject
to governance. Record every overlay at each reporting date, with the
reason the models do not capture the risk and the plan to bring the risk
into the models.

| Overlay ID | Reason | Segment | Amount (EUR million) *(Example)* | Method | Owner | Release condition |
|---|---|---|---|---|---|---|
| OVL-01 *(Example)* | Logistics property values falling faster than the commercial price index used in the LGD model | SEG-03 | 4.2 | LGD recomputed with a sector-specific 10% additional decline | Head of Credit Risk | New sector index integrated in the LGD model |
| OVL-02 *(Example)* | Energy-intensive SME borrowers not yet re-rated after a tariff increase | SEG-02 | 2.6 | Stage 2 transfer of identified borrowers, lifetime ECL | Credit Risk SME | Annual reviews completed |

The overlays are included in the 150.1 total. State their share of the
allowance (6.8 of 150.1, or 4.5%) and track it over time.

## 11. US GAAP CECL variant

<!-- Guidance: Complete this section for entities reporting under US GAAP (for example a US branch or subsidiary). Pitfall: copying the IFRS staging into CECL; CECL has no stages and measures lifetime losses from day one. -->

Under ASC Topic 326 (Subtopic 326-20 for financial assets at amortised
cost), the allowance reflects expected credit losses over the contractual
term, adjusted for expected prepayments, from initial recognition. The
entity forecasts over a reasonable and supportable period and then reverts
to historical loss information. Several methods are acceptable, including
loss-rate, vintage, roll-rate, discounted cash flow and PD x LGD methods.
ASU 2022-02 eliminated troubled debt restructuring accounting for
creditors and added disclosures of modifications made to borrowers
experiencing financial difficulty and of gross write-offs by vintage. The
interagency policy statement attached to SR 20-12 (revised 2023-04-21)
sets supervisory expectations. Check the codification for the current
text and effective dates that apply to your entity.

| Element | IFRS method | CECL adaptation *(Example)* |
|---|---|---|
| Horizon | 12 months or lifetime by stage | Lifetime for every loan |
| Forecast period | Explicit scenarios to 2028, reversion to anchor | Reasonable and supportable period 2 years, straight-line reversion over 1 year |
| Discounting | At EIR | PD x LGD method without discounting; DCF method for individually evaluated loans |
| Pools | Segments in section 3 | Same segments, US portfolio only |
| Individually evaluated | Stage 3 above threshold | Collateral-dependent loans at fair value of collateral less costs to sell |

*Example*: for the three-year loan in section 6.3, CECL would require a
lifetime estimate from origination, whatever its credit quality. For this
deteriorated loan that is about 24.33 thousand (discounted, or somewhat
higher undiscounted), against 11.14 thousand under IFRS 9 if it were
still in stage 1. For a loan with no deterioration, CECL would still
record a lifetime amount, computed with its lower lifetime PDs.

## 12. Validation, back-testing and monitoring

<!-- Guidance: Validate the models and the staging rules independently and at least annually, and monitor performance quarterly. Pitfall: validating the PD model but not the end-to-end ECL result. -->

Describe the validation scope (conceptual soundness, data, implementation,
outcomes), frequency and independence, and the monitoring indicators with
thresholds. Govern ECL models under Model Risk Management Policy
(`finance/model-risk-management-policy`) and record results in Model
Validation Report (`ai-data/model-validation-report`).

| Test ID | Test | Frequency | Threshold *(Example)* |
|---|---|---|---|
| VAL-T1 | PD back-test: predicted vs observed 12-month default rate by grade | Annual | Binomial test at 95% confidence, no more than 2 grades rejected |
| VAL-T2 | LGD back-test: predicted vs realised loss on closed defaults | Annual | Mean absolute difference below 5 percentage points |
| VAL-T3 | Stage 2 effectiveness: share of defaults that were in stage 2 beforehand | Quarterly | Above 70% |
| VAL-T4 | Stage stability: share of exposures moving stage more than twice a year | Quarterly | Below 5% |
| VAL-T5 | Allowance back-test: allowance at year start vs losses over the year in stage 1 | Annual | Within plus or minus 25% |
| VAL-T6 | Engine reconciliation: input exposures vs ledger | Quarterly | Difference below EUR 0.1 million |

Report validation findings to the ECL committee with a rating and a
remediation date, and decide whether an open finding needs a temporary
overlay until it is fixed. Between validations, the model owners monitor
the indicators above every quarter and explain any breach. A methodology
that has not been tested against realised losses for several years should
be treated as a key audit risk, whatever the size of the allowance.

## 13. Reporting and disclosure

<!-- Guidance: Produce the movement table and the stage tables from the engine, not by hand, and reconcile them to the ledger. Pitfall: transfers between stages shown net, which hides the real movement. -->

Prepare the disclosures required by IFRS 7 paragraphs 35A to 35N (credit
risk management, inputs and assumptions, reconciliations of the loss
allowance, credit risk exposure by grade) and the supervisory reporting
templates (for EU banks, FINREP). The reconciliation below is the core
table.

| Movement *(Example, EUR million)* | Stage 1 | Stage 2 | Stage 3 | Total |
|---|---|---|---|---|
| Opening allowance 2026-01-01 | [ ] | [ ] | [ ] | 132.4 |
| New originations | [ ] | [ ] | [ ] | 8.2 |
| Derecognitions and repayments | [ ] | [ ] | [ ] | (6.1) |
| Remeasurement, including stage transfers and model changes | [ ] | [ ] | [ ] | 21.3 |
| Write-offs | [ ] | [ ] | [ ] | (5.7) |
| **Closing allowance 2026-12-31** | **20.5** | **30.8** | **98.8** | **150.1** |

Check: 132.4 + 8.2 less 6.1 plus 21.3 less 5.7 = 150.1. Show transfers
between stages gross in your own table, and disclose the effect of
overlays, scenario weights and a sensitivity of the allowance to the key
judgements.

## 14. Glossary and references

<!-- Guidance: Use the same terms as the accounting policy and the regulatory reports. Pitfall: using "stage 3", "default", "non-performing" and "impaired" as if they always matched. -->

### 14.1 Glossary

| Term | Meaning in this methodology |
|---|---|
| CECL | Current expected credit loss model in US GAAP ASC Topic 326 |
| EAD | Exposure at default |
| ECL | Expected credit loss |
| EIR | Effective interest rate |
| LGD | Loss given default, as a share of EAD |
| Marginal PD | Probability of default in one period, unconditional on earlier survival |
| Overlay | Post-model adjustment approved by the ECL committee |
| PD | Probability of default |
| POCI | Purchased or originated credit-impaired asset |
| SICR | Significant increase in credit risk since initial recognition |
| Stage 1, 2, 3 | Industry terms for the IFRS 9 impairment buckets: performing, SICR, credit-impaired |

### 14.2 References

| Reference | Version or date checked | Use |
|---|---|---|
| IFRS 9 Financial Instruments, section 5.5, paragraphs 5.4.1, 5.4.3, 5.4.4, B5.5.37, Appendix A (IFRS Foundation) | Names and paragraph numbers only | Accounting requirements |
| IFRS 7 Financial Instruments: Disclosures, paragraphs 35A to 35N (IFRS Foundation) | Names and paragraph numbers only | Disclosures |
| FASB ASC Topic 326, Subtopic 326-20; ASU 2016-13; ASU 2022-02 | Names and numbers only | US GAAP CECL |
| EBA/GL/2017/06 Guidelines on credit institutions' credit risk management practices and accounting for expected credit losses | Published 2017-05-12; applies 2018-01-01 | Supervisory expectations |
| Basel Committee, Guidance on credit risk and accounting for expected credit losses (December 2015) | Name only | Background to EBA/GL/2017/06 |
| Regulation (EU) No 575/2013 Art. 178; EBA/GL/2016/07 (as amended by EBA/GL/2026/05, applying from 2026-10-19); Delegated Regulation (EU) 2018/171 | Article titles checked 2026-10-25 | Default definition |
| Federal Reserve SR 20-12, Interagency Policy Statement on Allowances for Credit Losses | 2020-05-08, revised 2023-04-21 | US supervisory expectations |

## Standard and licence note

Structure aligned with the impairment requirements of IFRS 9 Financial
Instruments and FASB ASC Topic 326, which are cited by name and paragraph
or topic number only; no text from either standard is reproduced or
paraphrased. IFRS is a trade mark of the IFRS Foundation, and this template
is not affiliated with or endorsed by the IFRS Foundation or the FASB.
Supervisory content is based on EBA/GL/2017/06 and EBA/GL/2016/07 (EBA
material may be reproduced with acknowledgement; no text is reproduced
here), Regulation (EU) No 575/2013 and the US interagency policy
statement attached to Federal Reserve SR 20-12 (a US government work). The
structure, wording, worked example and all figures are original;
Fernhollow Bank and Quillmoor Logistics GmbH are fictional. Check the
current official versions of every cited text before use. Template
licence: MIT.