Integration Capacity Analysis · matched-controls demonstration · June 2026
The warning that fired — and the one that didn't
Four collapses, four survivor twins, one instrument
Illustrative worked example · proof of concept · not a validated rating
Reading a collapse after the fact is easy. The test that matters is discrimination: does the
indicator drop the failure into the danger band while holding its look-alike survivor in the healthy band — under the
same shock, the same sector, the same year? Here the same structural indicator is run on four
famous failures, each paired with a near-twin that faced the same conditions and did not fail. Every
figure is computed only from information dated before the event.
How to read the score: a 0–1 structural-health indicator. Bands: Strong ≥ 0.80 · Stable 0.55–0.79 · Warning 0.30–0.54 · Critical < 0.30.
Status: illustrative — both sides chosen with the outcomes known. It shows the indicator can separate sound from hollow on transparent pre-event data; it is not a blind or out-of-sample result. The instrument used here was frozen by hash before scoring, and — separately — eight independent AI models reproduced this kind of read blind (see the Research update).
What this demonstrates — and what it does not
The claim. The same frozen indicator, computed only from pre-event facts, drops each casualty into Critical while holding its survivor-twin in Stable — under the same rate shock, the same industry, the same debt crisis. That is discrimination, not just consistency: the indicator does not merely label collapses after the fact, it stays quiet on the systems that looked similar but were sound.
The reel of collapses on its own proves little — four cases chosen because they failed will of course read as failing. The honest test is the pairing: for each failure, a near-twin that faced the same conditions and survived. If the indicator is real, it fires on one and stays quiet on the other. On four pairs, it does — and the one case where it partly fires on a survivor (Belgium) does so for a real, locatable reason, reported in full below rather than hidden.
The four pairs
1 · Banking — the 2023 rate shock
same rate environment, opposite structure
Silicon Valley Bank0.07Critical
+$1.5B FY22 net income · 15.26% CET1 · "well-capitalized"
Unrealized bond loss ≥ equity at 30 Sep 2022 · no Chief Risk Officer Apr'22–Jan'23 · 94% uninsured deposits
JPMorgan Chase0.65Stable
$37.7B net income · ROTCE 18% · CET1 13.2% · "fortress balance sheet"
CRO since 2013 · dedicated interest-rate-risk function · hedged / benefited from the rate rise
2 · Aviation — Q1 2019
same market, no safety-governance hollowing
Boeing 737 MAX0.085Critical
Record 2018 net income $10.46B · stock at an all-time high
MCAS single-sensor design · FAA-delegated self-certification · first crash signal not acted on
Airbus0.59Stable
€63.7B revenue · €3.05B net income · record 800 deliveries · backlog 7,577
A320neo ramp genuine · A380/A400M losses taken openly as charges
3 · Payments — 2020
same sector, real cash vs fabricated cash
Wirecard0.06–0.10Critical
Revenue €2.0B · in the DAX-30 · market cap > €24B
€1.9B "cash" never existed · auditor could not confirm the balances
Adyen0.72Stable
€496.7M net revenue (+42%) · €204M net income · 56% EBITDA margin
Free cash flow €259M (93% conversion) · same Big-Four auditor since 2009 · no restatements
Debt 96.7% of GDP (3rd-highest in EU) · Aa1 rating · full market access
No Eurostat reservation · 0.0pp data revision · legacy debt transparently managed
Honest partial-warning (recorded, not hidden): Belgium's governance channel flagged within months — a record 541-day government-formation deadlock (2010–2011), a genuine governing-capacity weakness. It did not threaten the data's integrity, solvency, or market access, and Belgium never defaulted. The indicator reads the real governing weakness Belgium had while still separating it sharply from Greece's solvency-and-disclosure collapse — evidence it is not tuned to wave survivors through.
Summary
Pair
Failure (score · band)
Survivor (score · band)
Separation
Banking 2023
SVB 0.07 · Critical
JPMorgan 0.65 · Stable
clean
Aviation 2019
Boeing 0.085 · Critical
Airbus 0.59 · Stable
clean
Payments 2020
Wirecard 0.06–0.10 · Critical
Adyen 0.72 · Stable
clean (strongest)
Sovereign 2010
Greece 0.10 · Critical
Belgium 0.55 · Stable (edge)
clean on solvency/data · partial governance warning
Across four matched pairs, the same frozen indicator put every failure in Critical and held every survivor in Stable, with an empty gap between the two bands (survivors 0.53–0.77; failures 0.02–0.23) — and it named the concealment mechanism each time (accounting classification, safety self-certification, fraud, statistical misreporting). That is discrimination on transparent pre-event data.
Limits
#
Limit
1
Illustrative, not validated. Both sides were chosen with the outcomes known. This shows the indicator can separate sound from hollow on transparent pre-event data; it does not establish a false-alarm rate. That needs blind, unfamiliar cases and the live forward register.
2
Semi-blind only. These cases are famous and so recognisable. The reproducibility result that backs them (eight independent models scoring blind) used the same cases — a fully blind version uses obscure systems the scorers cannot recognise.
3
Inputs distilled from the public record. Figures are drawn from filings, regulatory findings and Eurostat notifications, dated before each event; the structural pattern is what carries the reading, not the exact decimals.
4
Two checks pending before external use: a same-size regional-bank twin is a tighter match for SVB than JPMorgan; and the Airbus-vs-Boeing certification contrast should be cited to a European-regulator source.