# Credit Rating Rationale: [Obligor name], [Rating date]

> Internal credit rating file for one obligor and its facilities: rating
> summary, rating system and master scale, financial and qualitative
> model inputs, score-to-grade mapping, peer benchmarking, rating
> drivers, overrides with their justification, group and country
> considerations, facility LGD, CCF and EAD, outlook and re-rating
> triggers, data quality, rating history and independent approval.
> Based on the internal ratings-based (IRB) requirements of Regulation
> (EU) No 575/2013 as amended by Regulation (EU) 2024/1623 (CRR3),
> Articles 169 to 175 and 178, the EBA Guidelines on PD estimation, LGD
> estimation and treatment of defaulted assets (EBA/GL/2017/16) and the
> EBA Guidelines on loan origination and monitoring (EBA/GL/2020/06).
> Replace every `[bracketed]` value and every row marked *Example*.
> This is a document format, not financial, credit, regulatory or
> accounting advice; obtain the regulation and guidelines for the
> normative text.

<!--
Guidance: How to use this template
- One file per obligor rating event: new rating, annual review, event
  driven re-rating or default. The credit decision itself is in Credit
  Memo (`finance/credit-memo`); the yearly monitoring file is Annual
  Credit Review (`finance/annual-credit-review`). Both quote the grade
  this file approves.
- Banks with IRB permission must document rating assignments and
  overrides (CRR Articles 172(3) and 175). Banks on the standardised
  approach still use internal grades for pricing, limits and IFRS 9
  staging; keep the same discipline.
- This is a lender's internal rating, not a credit rating agency
  opinion. An investor's view of a bond issuer belongs in Bond Credit
  Analysis (`finance/bond-credit-analysis`).
- Fill sections 3 to 5 from the approved model only. Put every human
  judgement in section 8, never inside the model inputs.
- Worked example: fictional Fernhollow Bank re-rates the fictional
  Corvid Commerce Group at the annual review on 2027-05-20, using FY2026
  audited accounts (EUR thousand). Facilities: Fernhollow's 60% share of
  a EUR 50,000k club (Term EUR 18,000k drawn, Revolving EUR 12,000k
  undrawn), originated 2025-06-30 for three years and maturing
  2028-06-30, about 13 months after this rating date.
-->

## Document control

<!-- Guidance: The rater and the approver must be independent of the business that benefits from the credit (CRR Article 173(1)(a)). Record both. Pitfall: the relationship manager entering the rating and a colleague in the same front-office team approving it. -->

| Field | Value |
|---|---|
| Document ID | [RTG-YYYY-NNN] *Example:* RTG-2027-118 |
| Version | [1.0] |
| Status | [Draft / Submitted / Approved / Superseded] |
| Owner | [@credit-analyst] |
| Approvers/reviewers | [@credit-officer] (rating approver, second line), [@rating-desk] |
| Last updated | YYYY-MM-DD *Example:* 2027-05-20 |
| Classification | [Confidential: bank-internal risk information] |
| Obligor and customer ID | [Legal name, CIF] *Example:* Corvid Commerce Group plc, CIF-88412 |
| Group of connected clients | [GRP-NNNN] *Example:* GRP-2207 (rating applied to the borrower; guarantor subsidiaries are not direct obligors) |
| Rating event | [New / Annual review / Event-driven / Default] *Example:* Annual review |
| Rating model and version | [Model ID] *Example:* CORP-MID v4 (corporates with sales below EUR 500m) |
| Financial statements used | [Period, audited or not] *Example:* FY2026 audited consolidated, IFRS, approved 2027-03-10 |
| Final obligor grade | [Grade, PD] *Example:* FH-5, PD 0.60% |
| Next scheduled review | YYYY-MM-DD *Example:* 2028-05-20 (at least annually, CRR Art. 173(1)(b)) |

### Approvals

| Role | Name | Decision | Date |
|---|---|---|---|
| Credit analyst (rater) | [@name] | [Proposed] | YYYY-MM-DD |
| Rating approver (independent) | [@name] | [Approved / Returned] | YYYY-MM-DD |
| Override approver, if override applied | [@name] | [Approved / Rejected] | YYYY-MM-DD |

### Revision history

| Version | Date | Author | Change |
|---|---|---|---|
| 0.1 | YYYY-MM-DD | [@credit-analyst] | Model run and draft rationale |
| 0.2 | YYYY-MM-DD | [@credit-officer] | Override challenge and peer review |
| 1.0 | YYYY-MM-DD | [@credit-officer] | Approved and loaded in the rating system |

## Contents

1. [Rating summary](#1-rating-summary)
2. [Rating system and master scale](#2-rating-system-and-master-scale)
3. [Financial factors](#3-financial-factors)
4. [Qualitative factors](#4-qualitative-factors)
5. [Model output and grade mapping](#5-model-output-and-grade-mapping)
6. [Peer and benchmark comparison](#6-peer-and-benchmark-comparison)
7. [Rating drivers](#7-rating-drivers)
8. [Overrides and adjustments](#8-overrides-and-adjustments)
9. [Group, country and support considerations](#9-group-country-and-support-considerations)
10. [Facility rating: LGD, CCF and EAD](#10-facility-rating-lgd-ccf-and-ead)
11. [Outlook and re-rating triggers](#11-outlook-and-re-rating-triggers)
12. [Data quality and limitations](#12-data-quality-and-limitations)
13. [Rating history and performance tracking](#13-rating-history-and-performance-tracking)
14. [Glossary and references](#14-glossary-and-references)

## 1. Rating summary

<!-- Guidance: One table and one paragraph: model grade, override, final grade, PD, facility LGD, outlook and the two or three reasons. Pitfall: a summary that repeats the model output without saying why the final grade differs. -->

| Item | Result |
|---|---|
| Previous grade (date) | *Example:* FH-5, PD 0.60% (2026-05-22) |
| Model grade | *Example:* FH-4, PD 0.40% (weighted score 4.20) |
| Override | *Example:* One notch down (OVR-001) |
| **Final obligor grade** | **FH-5, PD 0.60%** *(Example)* |
| Facility LGD and EAD | *Example:* 32%; EUR 24,000k |
| One-year expected loss | *Example:* EUR 46k |
| Outlook | *Example:* Positive, conditional on refinancing signed by 2027-06-30 |
| IFRS 9 stage indication | *Example:* Stage 1 indication (grade improved since origination; formal SICR test on lifetime PD in the ECL engine) |

*Example:* The model improves Corvid by one grade because leverage fell to 0.62x and interest cover rose to 13.69x in FY2026. The approver confirmed a one-notch downward override: covenant EBITDA includes EUR 1,000k of net add-backs that flatter the margin input, and the FY2027 budget raises capex to EUR 12,000k (EUR 8,200k paid in FY2026) and the dividend to EUR 3,000k, so the forward cash cover is weaker than the historical inputs suggest. The grade stays FH-5 with a positive outlook; delivery of the FY2027 budget with DSCR above 1.60x would support FH-4 at the next review. The outlook is conditional: both facilities mature on 2028-06-30 and the borrower's board has mandated refinancing for signature by 2027-06-30, so failure to sign by then triggers an immediate re-rating (TRG-006).

Write the summary for a reader who knows the bank's rating system but not the borrower. Lead with the final grade and PD, then explain any difference from the model grade in one or two sentences. If the grade changed since the last review, say which inputs moved and by how much; a migration without a stated cause is the first thing an internal auditor or a supervisor will question. Keep the language factual: the rationale is evidence of an IRB process, and its tone should match that purpose.

## 2. Rating system and master scale

<!-- Guidance: Name the model, its scope and the master scale so that a reader can interpret the grade. CRR Article 170 requires a meaningful distribution of exposures across grades; Article 171(3) (as inserted by CRR3) requires a time horizon longer than one year and an assessment that holds under adverse conditions. Pitfall: using a model outside its approved scope (for example a mid-corporate model for a holding company or a project). -->

| Item | Detail |
|---|---|
| Model *(Example)* | CORP-MID v4: statistical financial module plus expert qualitative module, approved [date] |
| Scope | Non-financial corporates with consolidated sales of EUR 50m to EUR 500m |
| Large corporate check | Consolidated sales EUR 140,000k: not a large corporate under CRR Art. 142(1)(5a), so own LGD and CCF estimates may be used under the bank's permission |
| Rating philosophy | Hybrid, closer to through-the-cycle; PD calibrated to long-run average default rates with a margin of conservatism (EBA/GL/2017/16) |
| Default definition | CRR Art. 178 (90 days past due on a material obligation or unlikely to pay), applied per EBA/GL/2016/07 as amended |
| Model validation | Last independent validation [date]; see Model Validation Report (`ai-data/model-validation-report`) |

Master scale *(Example)*:

| Grade | Score band | PD (one year) | Category |
|---|---|---|---|
| FH-1 | below 1.75 | 0.05% | Performing, minimal risk |
| FH-2 | 1.75 to 2.49 | 0.10% | Performing |
| FH-3 | 2.50 to 3.24 | 0.20% | Performing |
| FH-4 | 3.25 to 4.24 | 0.40% | Performing |
| FH-5 | 4.25 to 5.24 | 0.60% | Performing |
| FH-6 | 5.25 to 6.24 | 0.85% | Performing |
| FH-7 | 6.25 to 7.24 | 1.50% | Performing, heightened attention |
| FH-8 | 7.25 to 8.24 | 3.00% | Watch list candidate |
| FH-9 | 8.25 to 9.24 | 7.00% | Watch list |
| FH-10 | 9.25 and above | 15.00% | Close monitoring, workout candidate |
| FH-D | n/a | 100% | Default |

The 0.05% PD of FH-1 equals the PD input floor for corporate exposures under CRR Article 160(1) as amended by CRR3; no performing grade may carry a lower PD in capital or expected loss calculations.

Explain the rating philosophy because it changes how a grade should be read. A point-in-time grade moves with the economic cycle and suits short-dated exposures and IFRS 9 inputs; a through-the-cycle grade is steadier and suits capital planning and limits. Most banks run a hybrid and convert grades for IFRS 9 using forward-looking adjustments. State which one applies so that readers do not compare grades across systems that use different philosophies.

## 3. Financial factors

<!-- Guidance: Enter the inputs exactly as the model defines them, from the stated financial statements, and show the calculation. Questions to answer: Which accounts were used? Were covenant or reported figures used, and does the model allow that? Pitfall: mixing management EBITDA in one ratio and covenant EBITDA in another. -->

| Factor | Weight | Definition | FY2026 value | Score (1 strong to 10 weak) | Weighted |
|---|---|---|---|---|---|
| Size *(Example)* | 10% | Consolidated revenue | EUR 140,000k | 7 | 0.70 |
| Leverage | 15% | Net debt incl. leases / EBITDA | 11,000 / 17,800 = 0.62x | 2 | 0.30 |
| Interest cover | 10% | EBITDA / finance charges | 17,800 / 1,300 = 13.69x | 2 | 0.20 |
| Cash flow to debt | 10% | (Cashflow - finance charges) / gross debt | (4,800 - 1,300) / 38,000 = 9.2% | 5 | 0.50 |
| Profitability | 10% | EBITDA margin | 17,800 / 140,000 = 12.7% | 5 | 0.50 |
| Liquidity | 5% | Cash plus undrawn committed lines vs 12-month needs | EUR 47,000k, ample | 3 | 0.15 |
| **Financial module** | **60%** | | | | **2.35** |

*Example:* The model specification uses covenant EBITDA as reported in the compliance certificate, because it is verified by the auditors on an agreed-upon procedures basis. Covenant EBITDA (EUR 17,800k) equals management EBITDA (EUR 16,800k) plus a restructuring add-back of EUR 900k and a share-based payment add-back of EUR 500k, less a disposal gain of EUR 400k. Section 8 tests the effect of using management EBITDA instead. Net debt of EUR 11,000k is the covenant figure: Term Facility 30,000 plus lease liabilities 8,000 (gross debt 38,000), less freely available cash of 27,000 (cash of 28,000 excluding 1,000 held in Vietnam).

Source each input from the identified statements, record the extraction date, and keep the spreading worksheet with the file. When accounts are qualified, restated or more than 15 months old, say so here and consider an override or a data quality flag.

Spread the statements in the bank's standard template before scoring, and reconcile the spreading to the audited totals: revenue, EBITDA, net debt and cash. Check the treatment of leases, capitalised development cost, factoring or supplier finance, and pension deficits, because each changes leverage and cash flow ratios. When the borrower changes its accounting presentation (for example under IFRS 18 from 2027), restate the comparatives on one basis so that the trend in the model inputs reflects performance rather than presentation.

## 4. Qualitative factors

<!-- Guidance: Score each factor against the model's written criteria and give evidence. EBA/GL/2020/06 paragraph 265 lists qualitative factors that affect repayment, such as management quality, owner commitment, pricing power and cost flexibility. Pitfall: scoring from memory of the relationship rather than from evidence dated within the review period. -->

| Factor | Weight | Score | Weighted | Evidence |
|---|---|---|---|---|
| Industry risk *(Example)* | 10% | 6 | 0.60 | Subscription share of revenue about 57% in FY2026 and rising; commerce remains cyclical |
| Market position | 10% | 5 | 0.50 | Growing share among small retailers in Germany and UK; small in the US |
| Management and governance | 5% | 4 | 0.20 | Delivered FY2025 and FY2026 above plan; independent chair; audit committee |
| Financial policy | 10% | 4 | 0.40 | Dividends within the covenant basket; leverage target below 1.5x stated by the board |
| Country and geography | 5% | 3 | 0.15 | Revenue mostly in Germany, UK and US; Vietnam small |
| **Qualitative module** | **40%** | | **1.85** | |

Write the evidence so that a second analyst would reach the same score. Where the model criteria do not fit the business (for example a platform business scored on a manufacturing grid), say so and consider a model scope exception rather than bending the scores.

Interviews with management, site visits and the relationship manager's knowledge are valid evidence, but record the date and the source. Where a score changed since the last review, give the reason in the evidence column, as in the example where the industry score improved because the recurring revenue share rose.

## 5. Model output and grade mapping

<!-- Guidance: Show the total score, the mapped grade and the PD. Pitfall: rounding the score to cross a band boundary. -->

$$
\text{Score} = \sum_{i} w_i \times s_i = 2.35 + 1.85 = 4.20
$$

*Example:* A weighted score of 4.20 falls in the FH-4 band (3.25 to 4.24), with a PD of 0.40%. The score sits 0.05 below the FH-5 boundary, so a small change in any input would move the grade: a one-point worse score on profitability adds 0.10 and gives 4.30, which maps to FH-5. Record any such boundary proximity because it is relevant to the override in section 8.

Run the model in the production rating system, not in a spreadsheet copy, so that the run date, model version and inputs are logged. If the system and a manual calculation disagree, the system result stands until the difference is explained and, where needed, corrected through change control. Attach the system output to this file.

| Comparison | Origination 2025-06-17 | Annual review 2026-05-22 | This review 2027-05-20 |
|---|---|---|---|
| Financial module *(Example)* | 3.00 | 2.70 | 2.35 |
| Qualitative module | 2.30 | 2.00 | 1.85 |
| Total score | 5.30 | 4.70 | 4.20 |
| Model grade | FH-6 | FH-5 | FH-4 |
| Final grade | FH-6 | FH-5 | FH-5 |

## 6. Peer and benchmark comparison

<!-- Guidance: Compare the obligor with similar rated obligors in the portfolio and with any external benchmark the bank uses. Pitfall: using peers from a different size band or business model, which makes the comparison meaningless. -->

| Peer (anonymised) | Sector | Revenue (EUR m) | Leverage | Interest cover | EBITDA margin | Grade |
|---|---|---|---|---|---|---|
| Corvid Commerce Group *(Example)* | Subscription commerce | 140 | 0.62x | 13.69x | 12.7% | FH-5 |
| Peer A *(Example)* | Retail software | 95 | 1.40x | 6.20x | 18.0% | FH-5 |
| Peer B *(Example)* | Online retail | 210 | 2.10x | 4.80x | 7.5% | FH-7 |
| Peer C *(Example)* | Logistics services | 180 | 1.10x | 9.00x | 10.5% | FH-5 |

*Example:* Corvid's credit metrics are stronger than Peer A, which shares the grade, but Peer A has a higher and steadier margin and a longer record. The comparison supports FH-5 rather than FH-4 at this stage. External benchmarks (for example, a market-implied PD from a credit data provider) can be added if the bank's policy uses them; record the source and date.

Choose peers from the same model segment and size band, using their latest approved ratings. Comment on any peer that carries a better grade with weaker metrics, because such inconsistencies can reveal either an outdated peer rating or a model weakness. Do not let a peer comparison replace the model: it is a challenge tool for the approver, and any resulting change must go through the override process in section 8.

## 7. Rating drivers

<!-- Guidance: Name the strengths and weaknesses that drive the grade, ranked by weight, so that triggers in section 11 follow naturally. Pitfall: listing ten drivers of equal importance. -->

| ID | Driver | Direction | Weight in decision |
|---|---|---|---|
| DRV-001 *(Example)* | Low leverage (0.62x) and high interest cover (13.69x) | Positive | High |
| DRV-002 *(Example)* | Recurring subscription revenue, about 57% of total | Positive | High |
| DRV-003 *(Example)* | Heavy capex absorbs most operating cash flow; DSCR 1.71x | Negative | High |
| DRV-004 *(Example)* | Short record of margin expansion (since FY2023) | Negative | Medium |
| DRV-005 *(Example)* | Bullet maturity 2028-06-30; refinancing mandated for signature by 2027-06-30 | Negative | High until signed |
| DRV-006 *(Example)* | Consumer exposure in commerce segment | Negative | Medium |

Explain in a short paragraph how the drivers combine. Ratings that rest on one strong metric deserve more caution than ratings supported by several independent strengths.

*Example:* The strong balance sheet (DRV-001) and recurring revenue (DRV-002) carry the grade. Against them, heavy capex (DRV-003) and the near maturity (DRV-005) limit it: at 2027-12-31 both facilities will have less than 12 months to run, the Term Facility becomes a current liability and the undrawn revolving facility stops counting as committed headroom in the borrower's treasury policy. With leverage at 0.62x the refinancing should be achievable, but until it is signed it is the main forward risk, which is why the approver kept FH-5 rather than accepting the model's FH-4.

## 8. Overrides and adjustments

<!-- Guidance: CRR Article 172(3) requires the bank to document when human judgement may override model inputs or outputs, who may approve overrides, every override made, and to analyse the later performance of overridden exposures. Give a reason that the model cannot capture, quantify it where possible, and stay within the override policy. Pitfall: overriding upward to keep a client in a pricing band. -->

| ID | Type | From | To | Reason | Approved by | Date |
|---|---|---|---|---|---|---|
| OVR-001 *(Example)* | Output override, downward | FH-4 | FH-5 | Covenant EBITDA add-backs flatter profitability; forward cash cover weaker than historical inputs; refinancing of the 2028-06-30 maturity not yet signed; score 0.05 from band boundary | [@credit-officer] | 2027-05-20 |
| OVR-002 | [Input override / Output override] | [Grade] | [Grade] | [Reason the model cannot capture] | [Authority] | YYYY-MM-DD |

Override policy *(Example)*: the analyst may propose up to one notch either way with the independent credit officer's approval; two notches need the head of credit risk; more than two notches, or any upward override above FH-4, need the rating committee. Upward overrides need evidence not available to the model, such as a binding guarantee or an equity injection after the reporting date.

### 8.1 Quantified support for OVR-001

*Example:* Replacing covenant EBITDA with management EBITDA (EUR 16,800k) moves the margin input to 12.0% and its score from 5 to 6, adding 0.10 to the total: 4.20 + 0.10 = 4.30, which maps to FH-5. The FY2027 budget increases capex from EUR 8,200k paid in FY2026 to EUR 12,000k and the dividend from EUR 2,600k to EUR 3,000k; budgeted cash flow is 21,600 - 2,100 - 3,300 - 12,000 = 4,200 against debt service of 1,200 + 1,600 = 2,800, a DSCR of 1.50x, down from 1.71x. Both points indicate that FH-4 would anticipate an improvement the borrower has not yet delivered.

Overrides are tracked in the override log, and the model owner analyses override performance by direction and by approver, as Article 172(3) requires; the bank's policy sets an annual analysis. A high override rate signals that the model needs recalibration; report it to the model risk function under Model Risk Management Policy (`finance/model-risk-management-policy`).

### 8.2 Questions the approver asks

Is the reason truly outside the model, or is it already captured by one of the inputs? Is the evidence dated and documented? Would the same reason lead to the same override for a comparable obligor? Does the override direction match the rating drivers in section 7? Is the override within the analyst's authority, and has it been entered in the override log with the approver's name? An override that cannot answer these questions should be rejected and the model grade kept, with the analyst's concern recorded as a monitoring point instead.

## 9. Group, country and support considerations

<!-- Guidance: State whether the grade is standalone or adjusted for parent or government support, and whether a country ceiling applies. CRR Article 172(1)(d) requires each separate legal entity to be rated; group policies must cover connected clients. Pitfall: notching up for parent support without a legal or economic basis. -->

| Item | Assessment |
|---|---|
| Rating level *(Example)* | Rating applied to the borrower, Corvid Commerce Group plc; guarantor subsidiaries are not direct obligors. Any subsidiary that borrows from the bank is rated separately (CRR Art. 172(1)(d)) |
| Parent or shareholder support | None assumed; the borrower is listed and no shareholder has any obligation to support it |
| Government support | Not applicable |
| Country ceiling | Not binding: parent and 90% of EBITDA in Germany, the UK and the US |
| Transfer and convertibility risk | Vietnam subsidiary cash (EUR 1,000k) not freely transferable; excluded from covenant cash |
| Connected clients | No other exposures in GRP-2207 |

Explain any notching between entities (for example, a weaker non-guarantor subsidiary rated below the group) and how intra-group dependencies were assessed. Where support is assumed, describe its form (guarantee, keep-well letter, strategic importance) and the evidence.

## 10. Facility rating: LGD, CCF and EAD

<!-- Guidance: CRR Article 172(1)(b) requires a facility grade at approval where the bank uses own LGD and CCF estimates. Show the LGD drivers, the downturn component and floors. Pitfall: reusing the LGD from origination after the guarantee or seniority has changed. -->

| Facility | Seniority and support | LGD (own estimate) | CCF | Limit (EUR thousand) | Drawn | EAD |
|---|---|---|---|---|---|---|
| FAC-001 Term Facility *(Example)* | Senior unsecured, guaranteed | 32% | n/a | 18,000 | 18,000 | 18,000 |
| FAC-002 Revolving Facility *(Example)* | Senior unsecured, guaranteed | 32% | 50% | 12,000 | 0 | 6,000 |
| **Total** | | | | **30,000** | **18,000** | **24,000** |

$$
\text{EL} = \text{PD} \times \text{LGD} \times \text{EAD} = 0.0060 \times 0.32 \times 24{,}000 = 46.1
$$

*Example:* The LGD model segments unsecured guaranteed corporate exposures by guarantor coverage and jurisdiction; it adds a downturn component estimated per EBA/GL/2017/16 and the bank's downturn LGD methodology. The 32% estimate exceeds the 25% LGD input floor for unsecured corporate exposures under CRR Article 161(4) as amended by CRR3. A large corporate would instead carry the foundation LGD of 40% (CRR Art. 161(1)(aa)). Counterparty exposure on the interest rate swap (notional EUR 12,600k) is measured separately under the bank's counterparty credit risk approach and is outside this table.

## 11. Outlook and re-rating triggers

<!-- Guidance: State the outlook and the specific, measurable events that would trigger an upgrade, a downgrade or an immediate re-rating. Article 173(1)(b) requires a new assignment when material information becomes available. Pitfall: vague triggers such as "significant deterioration". -->

| ID | Trigger | Threshold | Action |
|---|---|---|---|
| TRG-001 *(Example)* | Upgrade: FY2027 delivered | DSCR at or above 1.60x and management EBITDA margin at or above 13.0% | Consider FH-4 |
| TRG-002 *(Example)* | Downgrade: leverage | Covenant leverage above 1.50x for two quarters | Re-rate, expect FH-6 |
| TRG-003 *(Example)* | Downgrade: cash cover | DSCR below 1.35x at any test | Re-rate within 30 days |
| TRG-004 *(Example)* | Event: acquisition or dividend recapitalisation | Any debt-funded acquisition above EUR 10,000k | Immediate re-rating |
| TRG-005 *(Example)* | Event: covenant breach, late certificate, payment delay | Any | Immediate re-rating and watch list review |
| TRG-006 *(Example)* | Event: refinancing not signed | New facilities not signed by 2027-06-30 | Immediate re-rating; expect at least FH-6 and a negative outlook |

Link the triggers to the early warning indicators in the annual credit review so that monitoring and rating stay aligned.

Set triggers that the monitoring team can observe without re-running the full analysis: covenant ratios from the compliance certificate, payment behaviour, rating agency actions on related entities, and corporate events. Define the outlook in the bank's own terms (for example, positive means at least a one-in-three chance of an upgrade within 12 to 18 months) and use it consistently across the portfolio. A positive or negative outlook that persists for more than two reviews without a rating change should be challenged.

## 12. Data quality and limitations

<!-- Guidance: Record any weakness in the inputs and its effect. EBA/GL/2017/16 expects margins of conservatism for identified data deficiencies. Pitfall: leaving out known issues because the grade "looks right". -->

| ID | Issue | Effect | Treatment |
|---|---|---|---|
| DQ-001 *(Example)* | Covenant EBITDA used, includes add-backs | Margin input flattered by 0.7 points | Tested in OVR-001 |
| DQ-002 *(Example)* | Segment data unaudited | Industry score relies on management split | Accepted; reconciled to audited total |
| DQ-003 | [Issue] | [Effect] | [Treatment] |

State whether the model was used within its approved scope and whether any input was outside the range of the development data. A rating assigned with data more than 15 months old, or from unaudited accounts where audited ones are required, needs a flag in the rating system and approval at the next level.

## 13. Rating history and performance tracking

<!-- Guidance: Keep the full history so that migrations and override performance can be analysed. Pitfall: overwriting the previous record in the system without an audit trail. -->

| Date | Event | Model grade | Final grade | PD | Override | Approver |
|---|---|---|---|---|---|---|
| 2025-06-17 *(Example)* | Origination | FH-6 | FH-6 | 0.85% | None | [@credit-officer] |
| 2026-05-22 *(Example)* | Annual review | FH-5 | FH-5 | 0.60% | None | [@credit-officer] |
| 2027-05-20 *(Example)* | Annual review | FH-4 | FH-5 | 0.60% | OVR-001 (one notch down) | [@credit-officer] |

```mermaid
stateDiagram-v2
  [*] --> Draft
  Draft --> Submitted: analyst completes model run and rationale
  Submitted --> Returned: approver challenges
  Returned --> Submitted: analyst responds
  Submitted --> Approved: independent approval
  Approved --> UnderReview: annual cycle or trigger event
  UnderReview --> Submitted
  Approved --> [*]: obligor exits portfolio
```

Under IFRS 9, compare the current lifetime PD with the lifetime PD at initial recognition for the same remaining term to assess a significant increase in credit risk; the bank's criteria are in ECL Methodology (`finance/ecl-methodology`). The grade PDs in this file are one-year, through-the-cycle values, so they only indicate the direction. In the example the grade improved from FH-6 (0.85%) at origination to FH-5 (0.60%), which indicates Stage 1; the formal SICR test runs on point-in-time lifetime PDs in the ECL engine.

Retain every version of this file for the period the bank's record-keeping policy requires and at least as long as the data may be used for model development or validation. Rating histories feed default and migration statistics, PD back-testing and the override performance analysis, so missing or overwritten records weaken the whole rating system, not just one file.

## 14. Glossary and references

<!-- Guidance: Define each term once; list cited texts with version and date checked. -->

### 14.1 Glossary

| Term | Meaning |
|---|---|
| CCF | Credit conversion factor applied to undrawn commitments |
| Downturn LGD | LGD reflecting conditions of an economic downturn |
| EAD | Exposure at default |
| Facility grade | Grade or LGD class of a specific facility, as distinct from the obligor grade |
| Master scale | The bank's mapping of grades to PDs used across all rating models |
| Obligor grade | Grade expressing the borrower's risk of default |
| Override | Human judgement that changes a model input or output |
| PD input floor | Minimum PD that may be used in capital and expected loss formulae |
| Through-the-cycle | Rating approach that looks through short-term economic conditions |

### 14.2 References

| Reference | Version or date checked | Use |
|---|---|---|
| Regulation (EU) No 575/2013 (CRR), as amended by Regulation (EU) 2024/1623 (CRR3) | CRR3 OJ 2024-06-19, applying from 2025-01-01 (checked 2026-10-25); check the consolidated text | Arts 142(1)(5a), 160(1), 161(1)(aa), 161(4), 169 to 175, 178 |
| EBA Guidelines on PD estimation, LGD estimation and treatment of defaulted assets, EBA/GL/2017/16 | Applies from 2021-01-01 | PD calibration, margin of conservatism, downturn |
| EBA Guidelines on the application of the definition of default, EBA/GL/2016/07, as amended | Amended version applicable from 2026-10-19 (checked 2026-10-25) | Default definition |
| EBA Guidelines on loan origination and monitoring, EBA/GL/2020/06 | Applies from 2021-06-30 | Paragraphs 144(c), 257 to 265 |
| IFRS 9 Financial Instruments, impairment section 5.5 (IFRS Foundation) | Name and section number only | Staging reference |

## Standard and licence note

Based on Regulation (EU) No 575/2013 as amended by Regulation (EU)
2024/1623 and on EBA guidelines EBA/GL/2017/16, EBA/GL/2016/07 and
EBA/GL/2020/06 (EBA material may be reproduced with acknowledgement of
the source; none is reproduced here). Article and paragraph numbers are
cited and requirements are described in original wording. IFRS 9 is
cited by name and section number only; this template is not affiliated
with or endorsed by the EBA or the IFRS Foundation. The scorecard,
weights, master scale, PDs, LGDs and override policy in the example are
fictional illustrations, not regulatory values, except the cited PD and
LGD floors. Fernhollow Bank, Corvid Commerce Group and every figure are
fictional. This template is original text written for this collection.
Check the current official versions of the cited texts before use.
Template licence: MIT.