Across several EBA EU-wide stress test cycles for European G-SIBs, including the two most intensive of the past decade: the First-Time Adoption of IFRS 9 (2018) and the First-Time Adoption of CRR3 (2025). End-to-end: methodology design, credit/market/operational aggregation, supervisory submission, and live EBA challenge sessions.
Across several EBA EU-wide stress test exercises, the client — a European G-SIB with multi-risk trading and lending activity and significant ALM complexity — was always in the most demanding cohort.
Staging, SICR triggers, lifetime ECL and point-in-time PDs had to be re-engineered inside the stress engine, with credible downward migration through every scenario year. Nothing from prior IAS 39 cycles could be carried over.
Revised SA-CR for credit and counterparty risk, the Output Floor phase-in, reworked market risk rules and a redefined starting point. Legacy calibrations, peer benchmarks and RWA ratios all had to be rebuilt from scratch.
We took operational ownership of the stress test delivery, embedding 2 to 10+ senior advisors full-time within the client's Risk, Finance and ALM teams. We deployed the Ezelman ANCHOR™ framework — the seven-pillar protocol we use for every stress-testing starting-point, scenario design and reconciliation exercise.
Starting-point bridge, methodology workshops, scenario translation, IT infrastructure, reconciliation layer.
Full-scope projection (credit, market TB & BB, NII, op risk, P&L, capital), challenger models, template completion, submission.
EBA challenge sessions, written clarifications, findings-letter response, archive and handover to client.
No junior-only layers. Every deliverable signed by the senior who will defend it to the JST.
T-0 balance sheet assembled from 240+ counterparties, 180+ trading positions, 12 currencies — one source of truth.
IFRS vs prudential, solo vs consolidated, COREP vs FINREP — all reconciled before scenarios hit.
NPE and IFRS 9 stages locked at T-0; SICR triggers clearly defined for downward migration through scenarios.
Consistent calibration across credit, market, operational — no siloed assumptions, no double counting.
Clear accountability: who approves the T-0, who signs the scenario translation, who defends to EBA.
Three independent challenge rounds before submission. Every material assumption stress-tested internally.
Live EBA challenge sessions answered in real time — written clarifications within 24h for every query.
The ANCHOR pillars manifested in four concrete workstreams:
Rather than wait for the client's teams to build models, we ran parallel methodology workshops — credit, market, operational, liquidity — designing stress assumptions aligned with the EBA methodological note of each cycle (including the FTA IFRS 9 and FTA CRR3 editions). This allowed us to troubleshoot methodologies weeks ahead of client delivery, catching gaps early.
We designed the stress test "engine" — a unified framework that aggregated credit, market, and operational shocks, reconciled exposures across GL and BCBS 239 reporting, and produced auditable results. Every calculation was traceable to source data and assumption documentation.
We calibrated stress scenarios based on EBA guidance of each exercise and recent peer submissions, mapping macroeconomic drivers to balance sheet impacts. On the FTA IFRS 9 cycle, this meant re-engineering PD curves and staging logic inside the projection; on the FTA CRR3 cycle, it meant a clean-sheet view on SA-CR, Output Floor bite and the revised starting point.
Monthly validation reviews ensured methodology robustness. We prepared the bank for EBA challenge sessions by running live Q&A simulations, stress-testing the client's answers against likely supervisor questions.
Full-scope projection across every Pillar the EBA templates demand. Figures below are calibrated to the 2018 EU-wide stress-test transparency disclosures for the anonymised client cohort and cross-checked against EBA sector medians.
Stage-1/2/3 IFRS 9 staging and lifetime ECL re-engineered inside the engine; PD / LGD shocks across corporate, SME, retail and mortgage portfolios with PIT migration through every scenario year.
Rates, FX, equity and credit-spread shocks across the trading book with FRTB-aligned treatment by the 2025 cycle. Delta-weighted VaR, hedging effectiveness tests, sovereign revaluation overlay.
Loss-event frequency/severity calibrated to the bank's historical data and EBA peer benchmarks, with conduct-risk overlay (fines, remediation) aligned to supervisory trends.
Re-pricing curves, behavioural deposit caps, volume-constant rails under EBA methodology. Cross-currency book managed on a constant-balance-sheet basis through the 3-yr horizon.
Full P&L projection: NII, fees & commissions, trading income, credit losses, op-risk, taxes — aggregated on a single engine reconciled to the accounting close.
Full capital plan reconciliation: CET1 starting point, RWA migration, P&L impact, deductions and transitional filters. Well above the 4.5% CET1 regulatory minimum and 5.5% combined buffer-after-depletion test.
Every deliverable signed by a partner. Each shown with its complexity indicator so the Risk Committee can see where the work actually sits.
End-to-end credit stress test engine — fully auditable, with documented assumptions and data lineage. Same for all-risk consolidation process.
Four detailed macroeconomic scenarios (baseline, mild, severe, reverse) with supporting documentation of assumptions, peer benchmarks, and ECB alignment.
160-page technical document describing stress methodology, calibration rationale, model assumptions, and validation results. Pre-packaged for EBA submission.
Fully compliant EBA submission file (COREP format) with results across all scenarios and risk types. 15+ sensitivity analyses and reverse stress outputs.
Briefing documents and Q&A preparation for live EBA challenge sessions. Coached senior management on methodology explanations and assumption defences.
12-month plan for stress-test capability improvement, including data-quality upgrades and model enhancements for the next cycle.
The stress test was delivered on time and exceeded supervisory expectations. Each outcome shown below with its priority indicator.
Starting CET1 FL 11.8%. Baseline 2020 12.1%. Adverse 2020 8.64%. Well above the 4.5% CET1 regulatory minimum — demonstrated resilience through the full adverse scenario while clearing the combined-buffer hurdle.
Submitted prior to the EBA deadline, allowing time for pre-submission quality checks and final internal sign-off.
EBA challenge sessions resulted in 0 methodology critiques. Supervisors commended the transparency and auditability of assumptions.
Identified concentration risks in CIB leveraged finance and selected peripheral euro-area sovereign exposures, informing strategic risk decisions including balance-sheet re-weighting across CIB and retail lending.
Client Risk, Finance and ALM teams gained deep expertise in scenario design and stress aggregation, enabling independence on workstreams for future cycles.
Board Report on stress results became a flagship disclosure, improving investor confidence in the bank's risk-management capabilities.
A stress-test engine is only as defensible as the data that feeds it. Every Ezelman-built engine is shipped with a full lineage layer and a deployed set of data-quality controls, defaulting rules and completion algorithms — so the third-line audit review is a confirmation, not an investigation.
Credit core, trading book, ALM, GL, reference data. Each feed owned by a named source-system owner.
IFRS vs prudential, solo vs consolidated, COREP vs FINREP reconciled. Lineage tag attached to every record.
Credit, market (TB & BB), NII, op risk, sovereign. Calibration bench, not slide deck. Every parameter versioned.
Unified P&L, RWA and capital projection, reconciled to the accounting close and to the challenger model.
Submission-ready template with cell-level traceback to the source. No black-box steps.
Deployed at every stage: completeness checks, range validation, cross-system reconciliation, temporal consistency, ownership sign-off. Every breach logged, every exception justified or re-sourced before the value is allowed into the engine. No silent corrections.
Explicit, documented rules for missing values: source-system owner first, then reference-data master, then peer-cohort median, then conservative floor. Each rule versioned, approved by model-risk validation, applied deterministically — never discretionary. Every defaulted cell flagged in the audit log.
For structurally missing data (new portfolios without history, FTA-period gaps, merger-entity backfills) we deploy documented completion algorithms: regression on comparable cohort, time-series interpolation with regime-break detection, or supervisor-agreed proxy. Every algorithm walk-through is available on request.
No slide-ware, no associate intake. A direct, partner-level read on whether the same playbook applies to your situation — and whether you should hire us, hire someone else, or do it in-house.