Stress testing · ICAAP · Reverse stress testing

Calibrated from the calibration bench, not the slide deck.

Stress testing has two populations: banks that run it for the supervisor, and banks that run it for themselves and let the supervisor watch. The second is the one that holds Pillar 2G at zero. Ezelman engages only on the second register — reverse stress testing that is genuinely reverse, ICAAP that is genuinely capital-planning, EBA cycles that are genuinely defensible. None of our stress mandates are slide-deck mandates; all of them produce a model file and a narrative the JST can read.

0 bps
Median Pillar 2G add-on at SREP on Ezelman-run ICAAP cycles
10 wks
Typical duration of a reverse stress test done properly
3x
Depth of scenario library vs. standard EBA-cycle template
100%
Partner-led · no mandate delegated to a pyramid
Flagship mandate — EBA-wide cycle defence, European G-SIB, one cycle.CET1 peak-stress impact narrowed materially below peer-cohort median. Zero material supervisory follow-up on methodology.
Read the full case →

Qualitative outcome description, consistent with our sitewide policy. Peer-cohort reference derives from the EBA 2023 EU-wide stress test published results (aggregate CET1 depletion disclosed at bank level by the EBA); specific Ezelman basis-point outcomes are withheld at client request.

Ezelman PRISM™ — the stress-testing discipline

Five steps, one direction. Most stress cycles collapse at PRESSURE. Ours don’t.

Stress testing is not a simulation exercise. It is a risk-inventory exercise that happens to be expressed as a simulation. PRISM is the operating framework we apply to every stress mandate — EBA-cycle, SSM thematic, ICAAP, reverse, or ORSA adjacent. The order is not negotiable.

P
Step 1

PRESSURE

Catalogue the bank’s real pressure points, not the ones in the existing ICAAP. Concentrations, cross-entity dependencies, funding single-points-of-failure. The supervisor already has this map; the bank needs it too.

R
Step 2

REVERSE

A genuine reverse stress test: the scenario that breaks the bank, not the scenario that validates the current plan. If the exit scenario is not plausibly observable, the test is decorative. Most are.

I
Step 3

ITERATE

Calibration bench, not slide deck. Every scenario is iterated against parameter sensitivities, cross-correlations and model tails. Iteration is where hidden second-order effects get surfaced or buried — we surface them.

S
Step 4

SCENARIO

Translate scenarios into P&L, balance-sheet and RWA trajectories that the CFO recognises and the CRO can defend. Scenario narrative and quantitative impact are coupled. A scenario no one can tell is not a scenario.

M
Step 5

MAP

Map the stress result into ICAAP capital planning, management-action buffer logic, and the P2G dialogue. The stress cycle pays for itself only when it becomes the backbone of the capital conversation with the JST.

PRISM is the sequence; each step has a named deliverable, a named owner on the Ezelman side, and a named supervisor-facing artefact. The full PRISM deliverable tree is produced to the client steering committee at mandate inception.

Mandate types

Four stress-testing mandate patterns. All partner-led, all produce a model file.

Ezelman does not run “stress testing” as a monolith. The engagement shape differs structurally across the four patterns below — by team composition, by supervisory register, by deliverable cadence. The common thread is that every one produces a calibrated model the next cycle can inherit.

01

EBA-wide & SSM thematic cycles

Full-lifecycle defence of the bank’s submission to an EBA or SSM thematic stress test. Scenario translation, methodology interpretation, challenger modelling, explainability, the JST dialogue.

  • Scenario parameter translation into bank taxonomy
  • Challenger models on the largest 3 portfolios
  • Cross-check against peer-cohort disclosures
  • Result explainability dossier for the JST
  • Pre-submission dry-run with Risk Committee
02

Reverse stress testing (RST)

An RST that is actually reverse. Plausible break-point scenario, quantitatively calibrated, narratively defensible. Documented to the standard the ECB’s 2020 thematic review on RST governance set out.

  • Break-point identification workshop (CRO + CFO + Treasury)
  • Plausibility calibration against historical regimes
  • Management-action logic tree with trigger levels
  • Integration into ICAAP as a supplementary scenario
  • Annual refresh protocol
03

ICAAP & capital planning

ICAAP documentation the SREP letter will recognise — not a narrative overlay over last year’s numbers. Risk inventory, internal capital definition, buffer calibration, forward-looking capital plan.

  • Risk inventory refresh, stress-linked
  • Internal capital adequacy definition vs. regulatory CET1
  • Management buffer calibration against RST outcome
  • Forward-looking capital plan, 3-year horizon
  • Board-level presentation and challenge protocol
04

ORSA & cross-sector coupling

Where the client is a banking group with a material insurance arm, the ORSA and ICAAP are coupled. Ezelman runs the coupling exercise: consistent scenario library, cross-sector aggregation, unified management-action logic.

  • ORSA / ICAAP scenario alignment
  • Cross-sector capital aggregation methodology
  • Unified management-action framework
  • Group-level vs. solo-entity reconciliation
  • Dual supervisor dialogue (ECB + EIOPA or national)
Two recent mandates

What PRISM™ produces, in two real shapes.

Both anonymised. Both within the last 24 months. One was an EBA-wide defence; the other was a reverse stress test commissioned by the Board Risk Committee after a sharpened supervisory letter. The sequence was the same; the outcome was different because the starting pressure point was different.

Case A · EBA-wide cycle defence

French Tier One — EBA 2025 stress defence, 18 weeks, CET1 drawdown narrowed to under 2x peer median.

Starting position. A French Tier One entered the EBA 2025 cycle with a structurally diversified corporate-and-investment-banking book and a prior thematic finding on collateral recognition. Baseline internal modelling projected 430 bps of CET1 drawdown at peak stress — approximately 1.4x the peer cohort median.

PRISM sequence. Pressure-map surfaced three concentration tails the bank had not scenario-tested. Reverse-run calibrated a break-point at roughly 40% CRE haircut, plausibly observed twice in the last 30 years. Challenger models on the three tails; iteration loop narrowed divergence with bank’s model from 60 bps to 12 bps. Scenario narratives walked the JST through trigger logic step by step.

Outcome. Peak CET1 drawdown at submission narrowed materially below the peer-cohort median published by the EBA for the equivalent cycle. Zero material supervisory follow-up on methodology. Two remediation points on the prior thematic finding were closed at the same dialogue.
Qualitative outcome; specific basis-point figures withheld at client request. Peer-cohort reference: EBA 2023 EU-wide stress test published results.

Duration
18 weeks
Team
7 seniors, partner-led
Reg. register
EBA / SSM
Case B · Board-commissioned RST

Iberian mid-cap — reverse stress test after sharpened SREP letter, 10 weeks, 0 bps incremental P2G.

Starting position. An Iberian mid-cap bank received a SREP letter elevating concerns on “adequacy of reverse stress governance”. The Board Risk Committee asked for a genuine RST, externally calibrated, within the quarter.

PRISM sequence. Pressure map surfaced a cross-entity funding dependency the previous RST had not captured. Reverse-run exited on a plausibly observable wholesale-funding discontinuity. Iteration loop tested the trigger level against three historical regimes and two prospective-scenario overlays. Map step integrated the RST outcome into ICAAP as a named supplementary scenario.

Outcome. Next SREP cycle closed with zero incremental P2G add-on on stress governance. The Risk Committee Chair adopted the RST refresh protocol as an annual standing agenda item. No model remediation letter issued.

Duration
10 weeks
Team
4 seniors, partner-led
Reg. register
SSM · JST

Case A and Case B are real Ezelman mandates with identifying details anonymised for confidentiality. CET1 and Pillar 2 outcomes as reported by the client capital and risk functions at mandate close. Peer-cohort medians are Ezelman estimates from public EBA disclosures.

Why not the Big-4 for stress testing

Three structural reasons stress testing is a boutique problem, not a scale problem.

The Big-4 can staff an EBA-wide cycle. They can produce a deliverable that looks right. But three structural features of the stress-testing problem favour a partner-led boutique over a pyramid. The Risk Committee tends to discover this on the third cycle, not the first.

01 · Calibration depth

Stress testing is a modeller’s problem, not a methodologist’s problem.

The deliverable that matters is the calibration file, not the cover narrative. Ezelman’s named seniors sit at the calibration bench — they write the scenario parameterisation, not a PowerPoint summary of it. On a partner-led boutique the calibration quality is the mandate; on a pyramid, it is what juniors produce under review.

02 · Supervisory memory

The JST remembers the last cycle — and so should the advisor.

Stress-testing dialogue is multi-cycle. The JST tracks the bank’s methodology drift over three and four years. Ezelman runs consecutive cycles with the same partner lead, preserving supervisory memory. Big-4 rotation means a new senior interprets the prior cycle each time.

03 · Independence

We are not structurally conflicted on audit or vendor lines.

Ezelman does not hold a statutory audit practice. We do not resell, implement or earn partnership economics from any stress-testing software vendor. The calibration choice the Risk Committee receives is the one the model deserves — not the one that maximises the advisor’s downstream revenue. See the independence statement.

Where the Big-4 do win. When the bank needs a 20-country co-ordination on a thematic exercise that Ezelman’s capped roster cannot staff at pace, the Big-4 operating model delivers. We will say so and decline the mandate in those cases rather than run it thin.
Read-on · the long form

Three essays on the discipline behind the mandate.

Stress testing is not a slogan. Below are the essays the Risk Committee will want to read before commissioning an Ezelman-led cycle: on Pillar 2G formation, on reverse stress testing as Pillar 2 defence, and on top-down versus bottom-up scenario calibration.

Essay 01
Pillar 2G formation logic
How the supervisory machinery turns stress-testing weaknesses into capital add-ons — and the 18-month window banks have to influence it.
Essay 02
Reverse stress testing as a Pillar 2 defence
Why a genuinely reverse RST is the single highest-leverage artefact in the SREP dialogue — and what the ECB looks for.
Essay 03
Top-down vs bottom-up scenario calibration
Where top-down produces false comfort, where bottom-up produces unusable detail, and how PRISM™ reconciles them.

Before the next cycle begins.

Most banks engage on stress testing with eight weeks to go. The cycles that exit at zero P2G were scoped six months ahead. The 30-minute partner call is the place to understand which cycle you are in.

Book a partner call →
Follow Ezelman on LinkedIn — Regulatory insights read by 2,000+ risk professionals across Europe & the GCC Follow →