Every CRO in the eurozone has had this conversation. The SREP letter lands. Somewhere in the cover page, there is a new P2G number — a guidance line that is not legally binding, not publicly disclosed, and not in any formula the bank can reproduce. It is, in practice, a capital tax. The CFO opens the annex. The Group Treasurer opens the ICAAP policy. And then the programme office starts what we call the post-mortem reflex: "how did they arrive at this?"
By the time anyone is asking that question, the answer is already locked. P2G for the next SREP cycle is not being priced against the letter you just received — it is being priced against the ICAAP submission you are drafting right now, the stress-test run you closed three quarters ago, and the governance paper trail the JST has been quietly accumulating since the last on-site. The reason the advisory market rarely talks about P2G is that it is the single regulatory output where vendor dashboards, consultation papers and thought-leadership cycles cannot catch up with the real process. The formation window is slow. The signal is diffused. And the people who have sat on the other side of the table are not the people selling the model.
This essay is an attempt to put the machinery in one place. What P2G actually is. How a JST builds a candidate number. When that number becomes irreversible. And the operational test — not the modelling test — that separates banks exiting cycles at zero bps from banks that fund an extra thirty every year.
"Pillar 2R is modelled. Pillar 2G is formed. The distinction matters — because the machinery that forms it is governed by the bank, not the supervisor."
— From an Ezelman briefing, G-SIB CRO office, Q4 20251 · What P2G actually is — and what it isn't
Pillar 2 guidance sits between Pillar 2R (the binding capital requirement) and the combined buffer requirement. It is the ECB's stated view of how much CET1 a bank should hold above its total SREP capital requirement to absorb a severely adverse scenario without breaching its Maximum Distributable Amount trigger. It is not a formula. It is not, strictly, a rule. It is guidance that, if ignored, triggers a cascade of consequences: JST intensity, enhanced supervisory reporting, dividend conversations, and — in the current cycle — an active path toward Pillar 2R reclassification.
Two things to be precise about.
First, P2G is not buffer. It is not stacked on top of the combined buffer requirement in the MDA computation. It sits parallel, on its own trigger logic. The bank that plans capital as if P2G were part of the buffer stack is already misreading the letter.
Second, P2G is informed by the stress test but not determined by it. The EBA-ECB biennial stress test produces a projected CET1 depletion under adverse. That number feeds the P2G benchmark, but the JST can and does adjust upward for qualitative concerns — weak governance, unreliable stress-test infrastructure, or findings from an on-site. The stress test is the ceiling of a conversation, not the conclusion.
How P2G fits into the total capital demand
+ P2G = Overall Capital Expectation
(expectation ≠ requirement, but behaviour is identical)
The gap between requirement and expectation is the clean academic position. In practice, the market treats P2G as binding. Analysts model it as binding. Rating agencies reference it as binding. Boards manage to it as binding. The ECB knows this, and writes accordingly.
2 · The four inputs the JST is actually reading
The JST does not sit in a room with a calculator and pick a number. A P2G recommendation is built through a structured six- to nine-month conversation between the JST, the ECB's horizontal functions, and a small number of Directorate General analysts. Four inputs dominate that conversation. Three of them are entirely under bank control.
Stress-test projected CET1 depletion
The headline input. The ECB takes the bank's worst-year adverse CET1 ratio in the biennial EBA stress test — and the quality-assured ECB adjustments to it — as the anchor. A bank that depletes more than 300 bps in adverse is mechanically in P2G bucket territory. But the JST can override upward if the stress-test infrastructure itself fails a quality review — which is how a bank with a decent number still lands a painful add-on.
ICAAP credibility & internal stress scenarios
The ICAAP is read as a forecast of the bank's own honesty about its risk profile. The JST compares the bank's internal adverse to the EBA adverse. If the bank's internal scenario is materially milder — without a compelling business-model rationale — the JST treats that as underestimation and increases P2G. The banks that end up at zero bps run internal scenarios more severe than the EBA exercise, with documented reasoning.
Governance & Management-Body engagement
This is the soft input that banks systematically underweight. The JST records how the Management Body engages with the ICAAP, the stress test and the capital planning cycle. Evidence matters: observed board minutes, challenge questions, committee reviews, action-tracking. A bank whose MB rubber-stamps its ICAAP is signalling that the first and second lines are running capital with insufficient top-of-house scrutiny — and P2G moves up to compensate.
Post-on-site findings & horizontal reviews
Any OSI finding of Severity 3 or above in the two cycles preceding the SREP is a candidate P2G input — regardless of the ITR's direct capital impact. Horizontal reviews (Targeted Reviews of Internal Models, credit file deep-dives, IRRBB benchmarking) feed the same channel. The transmission is rarely one-for-one; it's cumulative. A cluster of findings in model governance plus an IRRBB benchmarking outlier plus a weak ICAAP can move P2G by 50 bps without any single input, alone, justifying it.
Three of the four inputs are governed by the bank — ICAAP, Management Body engagement, and how the bank responds to findings. Only the first is partially exogenous, and even there the quality of the underlying stress-test infrastructure is a bank decision. This is the first practitioner takeaway: P2G is not a supervisor output. It is a bank output scored by the supervisor.
3 · The bucket system — how a number becomes a letter
Since 2022 the ECB has used a structured P2G framework based on stress-test outputs, translated into four buckets of severity. The methodology was published, with deliberate opacity around the boundaries. The following table reflects the operational mapping we have seen applied across multiple SSM mandates — not the formal published ranges, but the working levels used in cycle modelling.
| Bucket | Stress-test CET1 depletion | Indicative P2G range | Typical JST posture |
|---|---|---|---|
| Bucket 1 | < 100 bps | 0 bps | Baseline — no add-on if ICAAP and governance align. |
| Bucket 2 | 100 – 250 bps | 100 – 150 bps | Starts to request written capital-planning responses; MB challenge quality closely read. |
| Bucket 3 | 250 – 400 bps | 150 – 250 bps | Full engagement; letter expected before draft SREP. Findings and ICAAP carry full weight. |
| Bucket 4 | > 400 bps | 250+ bps | Escalated; expect Director-level engagement, capital-plan review, P2R reclassification risk. |
Two practitioner notes on this table. First, the buckets are indicative anchors, not thresholds. The JST can land at 200 bps in Bucket 2 if governance is weak, or at 100 bps in Bucket 3 if the ICAAP narrative is exceptionally strong. The qualitative overlay is a 50–75 bps lever in either direction. Second, the ECB has indicated an appetite to re-tighten the boundaries in 2027 — specifically to narrow Bucket 1 and reduce the "free zone" that well-run banks have been using as a baseline. Planning for zero bps in 2028 will be measurably harder than in 2025.
Consider a Tier-1 bank with the stack that most European G-SIBs ran with into the 2024 SREP:
| Pillar 1 CET1 minimum (4.5%) + CCB (2.5%) + G-SII (1.0%) + CCyB (1.0%) | 9.00% |
| Pillar 2 Requirement (P2R, CET1 portion ≈ 56%) | 1.40% |
| MDA trigger (combined buffer requirement, CBR) | 10.40% |
| Pillar 2 Guidance (P2G) — pre-add-on | 0.00% |
| Reported CET1 ratio — Q4 2025 | 12.00% |
| Management buffer above MDA trigger | 160 bps |
The SREP letter lands with a 30 bps P2G. P2G does not hit the CBR — it sits above the MDA trigger as a supervisory expectation. Legally, the MDA distance is unchanged at 160 bps. Operationally, it collapses.
The operational MDA-distance maths — 12.00% − (10.40% + 0.30% P2G) = 130 bps. A 30 bps P2G consumes 18.75% of the management buffer on paper. But the Board's internal capital-deployment policy is typically calibrated to a "P2G-inclusive" management buffer — meaning the usable envelope for distribution, M&A or strategic RWA growth contracts by the full 30 bps. At a 14% payout ratio on a €600bn RWA G-SIB, that is €180m of foregone distribution capacity per annum, for as long as the add-on persists.
The practical takeaway: the P2G add-on never appears in the MDA formula, and never appears in Pillar 1 reporting. It sits in the Board pack as a silent tax on the buffer. Treasurers learn about it the first time a buy-back is deferred.
4 · The formation window — when the number is fixed
If there is one chart that should hang in the CRO's office, it is the formation window. The SREP letter arrives in Q4. The decisive window is Q1 to Q3 of the same year — and the anchor is set a year earlier. By the time the draft SREP is circulated internally at the ECB, the bank's influence on its own P2G is effectively zero.
"The draft number is usually set at JST-ECB horizontal triangulation, which is three months before the Supervisory Board sees it. The conversation banks think they're having in September was finished in June."
— Ezelman engagement debrief, 2024 SREP cycleThe practical sequence, stage by stage:
EBA stress test submission closes
The bank submits its templates. The ECB quality-assures. The anchor for the next cycle's P2G is functionally determined here. Everything downstream is overlay.
ICAAP submission & JST written assessment
The ICAAP is submitted; the JST produces a written supervisory assessment inside six to eight weeks. That assessment is the single largest qualitative input into P2G. It is also the input with the lowest bank-side attention.
Horizontal review & peer triangulation
The ECB's horizontal function compares the bank's stress-test and ICAAP outputs against peers. Outliers are flagged. This is when a mid-cap bank discovers that its IRRBB shock sensitivity is unusually mild — and the number starts to move.
JST recommends — Supervisory Board decides
The JST's recommendation goes to the Supervisory Board. The bank is not in the room. The Supervisory Board may adjust for system-wide considerations (e.g. raising all P2G by 25 bps if macro conditions warrant). At this point the number is fixed.
SREP letter delivered
The CRO reads the number. The advisory phone rings. Eighteen months too late.
5 · Where the money is
Thirty basis points of P2G on a €600 bn RWA G-SIB is €1.8 bn of CET1 locked up. At a 14% target ratio that is roughly 12% of the bank's entire CET1 stack — held, not deployed. The opportunity cost at a 10% ROE target is €180 m of foregone earnings per year, for as long as the add-on persists. Across a three-cycle memory window, which is the standard supervisory dwell time for a P2G number, the total earnings cost is closer to €540 m.
Compared to that, the cost of a ten-month remediation programme — governance re-papering, ICAAP rebuild, stress-test infrastructure upgrade, and on-site close-out — is a rounding error. This is the economic argument we make on every mandate. Not that a P2G add-on is unfair, or avoidable, or a failure of the JST. The argument is simpler: the capital cost of one cycle of mispriced P2G pays for three cycles of doing it properly.
6 · The operational test — seven questions that move the number
Most CROs we speak to want a diagnostic, not a framework. Here is one. If the answers to all seven are "yes, with documented evidence," the bank is operationally positioned for zero-bps exit. Three or more "no" answers correlate, in our experience, with an illustrative 50–150 bps add-on in the following cycle — a range consistent with the P2G bucket taxonomy published by the ECB SSM in the annual SREP methodology booklet.
Seven operational tests · run them two quarters before ICAAP submission
- Does your internal adverse scenario produce a CET1 depletion equal to or greater than the last EBA adverse, with documented, bank-specific rationale where it diverges?
- Is your ICAAP narrative structured so that each material risk has an explicit capital-impact paragraph that a non-technical reader can trace to the stress-test output?
- Can the Management Body demonstrate, in minute extracts, at least four substantive challenge questions on the ICAAP in the preceding twelve months — not procedural ones?
- Is every open Severity-3+ OSI finding mapped to a capital-planning mitigation with a board-signed closure date, and is progress reported monthly to the CRO?
- Does your IRRBB and credit concentration benchmarking place the bank within a justifiable band versus peers — and if it doesn't, do you have a proactive JST briefing explaining why?
- Is your stress-test infrastructure documented to the standard of a Targeted Review of Internal Models submission — reproducibility, data lineage, model validation coverage, challenger model?
- Does your CRO own the JST relationship end-to-end — with monthly technical contact between SREP letters — rather than delegating it to a regulatory-affairs function?
7 · What the advisory market won't tell you
The reason Pillar 2G receives less consulting attention than Pillar 1 or Pillar 2R is mechanical. Pillar 1 is a data-and-models exercise, which is where systems integrators earn. Pillar 2R is a modelling exercise, which is where the quant desks at the Big Four earn. Pillar 2G is a judgement exercise — formed at the intersection of stress-test infrastructure, governance papering, MB engagement and JST relationship management. None of those four activities is billable on a day rate at volume.
That does not make the work less addressable. It makes it harder to package. The firms that understand how to reduce P2G are not the firms that build the model — they are the firms that have sat on the JST side of the conversation, or on the Risk Committee side of it, often both. The question to ask any adviser pitching you on "SREP readiness" is not whether they have a framework. It is whether they have reduced a real P2G number, in a real SSM cycle, and whether they will name it on a reference call.
If your answer to that is no: the programme you are buying is unlikely to move the number.
8 · Closing — the long view
Pillar 2G is a slow signal. It forms over eighteen months, it anchors three cycles, and it costs the bank more CET1, over time, than almost any other supervisory output. It is also, uniquely in the Pillar 2 framework, reversible — a well-run cycle can walk a bank from 100 bps to zero in a single SREP if the infrastructure, governance and JST relationship are rebuilt together.
The banks that do it right are not smarter than their peers. They are simply earlier. They rebuild the ICAAP six months before submission, not six weeks. They rehearse the MB challenge. They run the internal adverse harder than the EBA adverse. They close OSI findings on a board-signed timeline, not a remediation plan. They treat the JST relationship as a monthly calendar, not an annual one. None of this is expensive. All of it is hard to retrofit.
The single sentence we leave with every CFO and CRO we brief: the cheapest P2G is the one you don't pay. If you are inside the formation window now, this is the quarter the number moves. If you are not, the next one is twelve months out — and the board conversation that matters should have started yesterday.