The EBA’s consultation on revised stress test templates demands a fundamental rethink of data architecture, projection engines, and quality assurance frameworks. Banks that treat this as a reporting change will fail.
Side-by-side comparison of old vs. new template requirements across each major reporting area, with impact severity based on implementation effort and data infrastructure demands.
| Reporting Area | Old Template | New Template | Key Change | Impact |
|---|---|---|---|---|
| Credit Risk — Losses | Portfolio-level aggregation, 15 segments | Exposure-level granularity, 45+ segments | 3x segmentation depth; counterparty-level data | CRITICAL |
| NII Projections | Top-down NII with 5 repricing buckets | Bottom-up NII with 12 repricing buckets + behavioural modelling | Granular repricing; requires loan-level cash flow engine | CRITICAL |
| Market Risk P&L | Aggregated P&L by risk factor | P&L attribution by desk, risk factor, and instrument type | Desk-level attribution; reconciliation to FRTB sensitivities | HIGH |
| Operational Risk | Scenario-based aggregate loss | Scenario-based loss + frequency/severity decomposition | Requires granular OpRisk scenario library | HIGH |
| Sovereign Exposures | Country-level aggregation | Instrument-level with maturity + accounting classification | HTCS/HTC split required; mark-to-market detail | MEDIUM |
| Capital Ratios | Point-in-time CET1/T1/TC | Dynamic capital with management action constraints | Standardised management action framework | MEDIUM |
| Liquidity & Funding | LCR/NSFR snapshot | LCR/NSFR + cash flow survival analysis under stress | New template section; requires liquidity stress engine | HIGH |
Assessment of data readiness across risk types and data dimensions. Red indicates significant gaps requiring new data sourcing; amber indicates partial coverage needing enrichment; green indicates existing infrastructure that is adequate.
| Data Dimension | Credit Risk | Market Risk | Operational Risk | NII / ALM | Liquidity |
|---|---|---|---|---|---|
| Counterparty-Level Attributes | AMBER | GREEN | RED | AMBER | RED |
| Exposure-Level Granularity | RED | AMBER | RED | RED | AMBER |
| Historical Time Series (7yr+) | AMBER | GREEN | AMBER | AMBER | RED |
| Collateral & Recovery Data | RED | GREEN | GREEN | GREEN | GREEN |
| Behavioural Model Inputs | AMBER | AMBER | RED | RED | RED |
| Macro-Variable Linkage | AMBER | AMBER | AMBER | AMBER | AMBER |
Exposure-level granularity across credit risk and NII is the single largest infrastructure challenge. Most banks’ stress testing platforms were built on portfolio-aggregate data. Retrofitting exposure-level data pipelines is a 12–18 month undertaking that requires concurrent validation.
Each risk type faces specific methodology changes in the revised framework. Banks must update projection engines, recalibrate models, and validate outputs against supervisory benchmarks.
PD/LGD satellite models must now incorporate forward-looking macroeconomic variables with documented causal mechanisms. Stage migration under IFRS 9 must be modelled dynamically, not via static transition matrices. Sector-specific loss rate projections required for CRE, leveraged lending, and consumer unsecured.
Desk-level P&L attribution replaces aggregate risk factor decomposition. Requires reconciliation between accounting P&L and risk P&L under stress. New sensitivity-based approach aligns with FRTB reporting. Banks must demonstrate that stressed P&L is consistent with their IMA/SA-TB calculations.
Scenario-based projections must now decompose into frequency and severity components with explicit distribution assumptions. Banks must maintain a minimum of 15 operational risk scenarios covering cyber, conduct, legal, and business continuity. External loss data integration is mandatory.
Bottom-up NII projection replaces top-down approaches. Loan-level cash flow modelling with behavioural repricing assumptions (prepayment, drawdown, deposit migration) required. Net fee income must be stressed separately with volume and margin sensitivities linked to macroeconomic scenarios.
The EBA mandates a structured QA process that goes far beyond automated validation checks. Banks must demonstrate independent challenge, supervisory benchmark comparison, and documented governance sign-off at each stage.
Automated cross-template reconciliation, arithmetic consistency checks, sign tests, and time-series plausibility filters. Must cover 100% of submitted data fields. Errors caught at this tier should be zero by submission date — the EBA considers Tier 1 failures as indicative of weak processes.
Independent review of projection methodology, key assumptions, and scenario application by second-line risk functions. Must document challenges raised, management responses, and resolution outcomes. The EBA expects evidence that experts questioned aggressive assumptions and that disagreements were escalated.
Banks must compare their projections against EBA/ECB benchmark models and explain material deviations. Deviations >20% from supervisory benchmarks require documented justification with supporting analysis. Unexplained deviations trigger supervisory follow-up and potential top-down adjustments.
The three-tier QA framework requires 4–6 weeks of calendar time between first draft results and final submission. Banks that compress QA into the last week before submission produce visibly lower-quality outputs that attract supervisory scrutiny and top-down adjustments.
The consultation period is your window to influence the final framework. Generic responses are ignored. Evidence-backed, technically specific feedback on these five areas will shape the final text.
Argue for tiered granularity requirements based on balance sheet size and complexity. Provide concrete evidence: how many FTEs and months would exposure-level credit risk reporting require for a bank with <30 IRB models? The EBA has signalled openness to proportionality if backed by data.
Document your current end-to-end process timeline (scenario receipt to submission) and demonstrate that a 40% compression is unrealistic without quality trade-offs. Propose specific alternative milestones with quality checkpoints that maintain supervisory confidence.
Push for clarity on acceptable behavioural assumption frameworks for NII. The current draft leaves too much ambiguity on deposit migration, prepayment modelling, and repricing lag assumptions. Request supervisory guidance or safe-harbour parameters.
Advocate for realistic management action assumptions. The proposed constraints on balance sheet optimisation, pricing responses, and portfolio de-risking may force banks to project unrealistic static balance sheets. Provide evidence from past crises showing actual management responses.
Request that the EBA publish benchmark model specifications and calibrations before the exercise starts. Banks cannot meaningfully compare their projections to benchmarks if the benchmark methodology is opaque. Transparent benchmarks enable better self-assessment.
Request a phased data quality standard with a dry run using relaxed completeness requirements before the full exercise. This allows banks to identify and remediate data gaps without the pressure of supervisory submission deadlines on the first attempt.
The path from consultation response to operational readiness requires disciplined milestone management. Each phase has dependencies that create knock-on delays if missed.
Can your bank build exposure-level data pipelines, recalibrate projection engines, implement three-tier QA, and run a dry run — all within 18 months? For most banks, the answer is no without starting infrastructure work now, before the final text is published.
A phased approach with resource estimates for each workstream. Banks should initiate Phase 1 immediately — waiting for the final text to start infrastructure work guarantees timeline failure.
Full programme delivery requires 15–25 FTE across risk methodology, data engineering, IT, and programme management over 18 months. Total investment for a G-SIB: €5–12m including technology, external support, and internal resource costs. Under-investment creates execution risk that materialises as supervisory top-down adjustments.