The most difficult ECL audit questions are rarely answered by the final allowance alone. Reviewers want to understand how the population was assembled, why exposures changed stage, which assumptions changed, where management judgement entered and how the approved result reached the financial statements.
Preparing the following ten working papers as part of the normal close creates a cleaner review path and reduces retrospective evidence gathering.
1. Scope and portfolio inventory
Start with an inventory of financial assets and off-balance-sheet exposures considered for impairment. Identify the accounting category, measurement approach, materiality, source system, owner and any exclusions.
The paper should reconcile the ECL scope to relevant balance-sheet and off-balance-sheet records and explain items assessed individually or under a different method.
Include: scope decision, portfolio totals, route to measurement, exclusions and approval.
2. Source-data reconciliation and quality report
Reconcile extracted balances and counts to subledgers, trial balance and other controlled sources. Summarise data validation, missing fields, duplicates, late feeds and material exceptions.
Do not provide only a clean final file. Preserve the original control totals, detected exceptions, corrections and reviewer sign-off so the process can be followed.
Include: source totals, model totals, reconciling items, exception treatment and sign-off.
3. Segmentation and methodology paper
Map each portfolio to its ECL method and explain why the method and segments fit the underlying risk. Summarise changes from the prior period, data limitations and any newly introduced products.
For receivables, this may cover provision-matrix segmentation. For loans, it may explain PD–LGD–EAD components, individual assessment and treatment of commitments.
Include: methodology map, segment definitions, rationale, changes, limitations and approval.
4. SICR, stage movement and override report
Show opening and closing Stage 1, Stage 2 and Stage 3 balances and allowances, together with transfers. Explain material migration by portfolio and include quantitative triggers, qualitative indicators, backstops, cure and default definitions.
List manual overrides with original stage, revised stage, rationale, evidence and approver. Investigate direct Stage 1-to-Stage 3 movement and unusual cure patterns.
Include: movement matrix, trigger summary, override register, exceptions and management commentary.
5. Model and parameter summary
Identify every production model and parameter set used in the period: PD, LGD, EAD, provision rates, transition matrices, recovery assumptions, credit conversion factors, effective interest rates and other key inputs.
The report should distinguish unchanged values from refreshed, recalibrated or manually adjusted values and link each change to testing and approval.
Include: version, effective date, owner, change reason, impact and approval.
6. Forward-looking scenarios and sensitivity paper
Document economic scenarios, variables, forecast sources, horizons, weights, model linkages and post-forecast treatment. Explain changes from the prior period and consistency with other internal planning views.
Sensitivities should show how ECL responds to alternative weights or assumptions. Their purpose is not to select the most convenient answer but to reveal estimation uncertainty and dominant drivers.
Include: scenario paths, weights, rationale, sensitivity, challenge and approval.
7. Overlay and post-model adjustment memorandum
Maintain one inventory of all overlays, including unchanged, new, revised and released items. Each memo should identify the uncovered risk, targeted population, methodology, amount, supporting indicators, double-counting assessment and exit criteria.
Reviewers should be able to understand why the model output was insufficient and why the chosen adjustment is proportionate.
Include: calculation, evidence, alternatives, sensitivity, governance, expiry and outcome monitoring.
8. Validation, backtesting and monitoring report
Summarise independent validation status, open findings and interim controls. Include current performance monitoring for discrimination, calibration, realised loss, recovery, stage effectiveness, population stability and data quality as relevant.
If a model is used with an overdue finding, explain management acceptance, estimated impact, compensating control and remediation date.
Include: test results, thresholds, breaches, findings, owners, due dates and closure evidence.
9. ECL movement bridge and journal-entry pack
Bridge opening to closing ECL through understandable drivers: new originations, repayments, derecognition, stage transfer, parameter refresh, scenario change, overlays, write-offs, recoveries and other effects.
Reconcile the final approved result to the general ledger and preserve the exact file/version from which the journal was prepared. Include journal approval and posting evidence.
Include: movement bridge, model-to-ledger tie-out, journal, approval and posting confirmation.
10. Disclosure tie-out and governance pack
Tie quantitative disclosures to approved ECL outputs and reconcile qualitative statements to the period's actual methodology and judgements. Include allowance reconciliation, credit-risk grades or delinquency information, write-offs, modifications, collateral and relevant assumption disclosures.
Complete the pack with committee papers, minutes, challenge, decisions and final approval. The narrative should agree with what governance forums actually considered.
Include: disclosure checklist, tie-out, narrative support, committee record and final sign-off.
Suggested folder structure
ECL_YYYY_MM/
01_scope_and_inventory/
02_data_and_reconciliations/
03_methodology_and_segmentation/
04_sicr_staging_and_overrides/
05_models_and_parameters/
06_scenarios_and_sensitivities/
07_overlays/
08_validation_and_monitoring/
09_movement_and_journal/
10_disclosures_and_governance/
final_run_manifest.mdAccess, naming and retention should follow the institution's policies. The essential principle is that the period can be reconstructed without relying on one analyst's inbox or memory.
Common documentation mistakes
- Papers describe policy but not what occurred in the reporting period.
- Evidence is created after the reviewer asks for it.
- Draft and final outputs cannot be distinguished.
- Reconciliations show a difference but not its resolution.
- Committee minutes record approval without challenge or rationale.
- Overlay support explains the amount but not its release conditions.
- Disclosures are prepared from a separate spreadsheet that does not tie to the booked result.
Frequently asked questions
Should every paper be lengthy?
No. Documentation should be proportionate to materiality, complexity and judgement. A concise controlled report can be stronger than a long narrative if it clearly states the conclusion, evidence, reviewer and changes.
Can system reports replace working papers?
They can provide much of the evidence if they capture the relevant population, logic, version and approval. Narrative judgement—especially scenarios, overlays and limitations—often still needs a governed memorandum or workflow record.
Who should own the audit pack?
Ownership is usually shared across finance, risk, modelling and data teams, with one accountable coordinator for completeness and final approval. Individual papers should name preparers and reviewers.
Make audit readiness part of the close
Review the Audit-Ready ECL Reporting Software workflow or explore Controls, Documentation and Audit Readiness.
Technical references
- IAASB, Auditing Accounting Estimates—ISA 540 (Revised)
- IFRS Foundation, IFRS 9 educational session: disclosures