DEBT beyond code.
The four axes were written for code concepts. But Detectability, Effort, Burden and Trap are really questions about the cost of being wrong — and that question exists everywhere money and decisions do. These cases show the axes at work in finance, where the stakes are public and the post-mortems are brutal.
A note on sourcing: every case below is real and recent. None is named. The companies are hinted at deliberately — the point of a case study is the pattern, not the scalp. If you follow financial press, you'll recognise them in a sentence or two. If you don't, the lesson works anyway.
Case 1 — The conglomerate nobody could read
Picture a founder-controlled automotive-parts conglomerate assembled through years of debt-funded acquisitions. Growth looked relentless. The financing behind it ran partly through off-balance-sheet vehicles. Information requests from outside parties were met with reluctance; the founder's history carried old fraud allegations that diligence teams flagged and the market shrugged off.
Then it collapsed into bankruptcy — and the striking part is who saw it coming, and how early. A major rating agency's machine-readable probability-of-default model reportedly flagged acute stress a full ten months before the filing. The signal existed. It was scorable. Most investors simply weren't consuming risk in that form — they were reading narratives, and the narrative was opaque by design.
- D9 — Detectability
- The core failure. Off-balance-sheet structures, founder opacity and refused disclosure pushed detection from "read the filings" to "forensic audit." When the truth requires forensics, almost nobody does the forensics.
- E8 — Effort
- A distressed-debt fund's red-flag methodology captures this exactly: when a counterparty refuses basic transparency, the effort to remediate any underlying issue is exponentially higher. You can't fix what you're not allowed to see.
- B8 — Burden
- Serial debt-funded M&A is a one-way architectural commitment. Every acquisition deepened the structure that made unwinding impossible.
- T6 — Trap
- The growth story felt like success. Relentless expansion reads as strength right up until it reads as leverage.
d1 is "visible in the diff," d9 is "requires forensics." The rule of thumb: if your risk is only visible to forensic tooling, assume nobody is looking.
Case 2 — The metric lenders trust because it can't be argued with
For decades, commercial real-estate lending leaned on Loan-to-Value. LTV has a quiet flaw: the "V" is an appraisal, and appraisals are opinions. Opinions inflate in good times, which is precisely when you need them not to.
European rating analysts covering the sector now push a second number alongside it: Debt Yield — net operating income divided by the loan. No appraisal in the formula. Their language for why is almost a Burden-axis definition: the metric is hard to manipulate, dynamic (it moves when cashflow moves, not when sentiment moves), and appropriate for peer analysis because every property is scored against the same hard data.
- T8 — Trap
- Appraisal-anchored LTV is a textbook trap: it looks rigorous — a ratio! with decimals! — while quietly importing a subjective input. The format of the number hides the softness of its source.
- B8 — Burden
- Loans underwritten on inflated values are long-term commitments to a wrong number. The weight lands years later, at refinancing, when the appraisal cycle has turned and the cashflow hasn't grown into the loan.
- D6 / E3
- Detecting an inflated appraisal mid-cycle is genuinely hard. The fix, once the industry named the problem, was cheap: add one manipulation-resistant metric to the underwriting sheet.
b_score anchors encode for code concepts: measure the long-term weight of choosing wrong using data that can't be negotiated with. A good metric is one nobody can flatter.
Case 3 — The airport that can only pay in cash
Take a major European airport operator — a household name in its home market, infrastructure that isn't going anywhere. The income statement shows profits. Equity analysis flags it as risky anyway: net debt running at 7.4× EBITDA, and free cash flow going the wrong direction while capital projects burn.
The analysts' conclusion is the kind of sentence DEBT exists to formalise: a company can only pay off debt with cold hard cash, not accounting profits. Paper profitability and the actual capacity to service obligations are two different measurements, and only one of them retires debt.
- E7 — Effort
- True remediation cost here is paid in cash: selling assets, cutting capex, raising equity. None of it is an accounting entry. The gap between "fix it on paper" and "fix it for real" is the whole Effort axis.
- B8 — Burden
- 7.4× leverage on long-lived infrastructure is a decade-scale commitment. Every year it persists, the option space narrows.
- T8 — Trap
- "It's profitable" is the seductive misread. Accounting profit feels like solvency the same way passing tests feels like correctness.
- D5 — Detectability
- Middling on purpose: the numbers were public. Anyone could compute debt-to-EBITDA. The signal wasn't hidden — it was unfashionable.
What the three cases share
Strip the sector away and the same structure repeats:
- The risk was real and the data existed. In every case, a quantitative reading was available before the damage — a default-probability model, a manipulation-resistant ratio, a public leverage multiple.
- The failure was a missing axis, not missing intelligence. Smart people looked at each situation. They were scoring on one dimension while the danger lived on another.
- Top practitioners already think in multiple axes — intuitively. The rating agency's model, the debt-yield push, the red-flag diligence checklists: each is one axis of DEBT, rebuilt from scratch, in isolation, without a shared vocabulary.
So is DEBT a finance standard now?
No — and it's important to say that plainly. The published anchors (spec) are written for code concepts, calibrated against code examples, and that's what the scores in the wild mean. A b8 in the corpus is "architectural one-way door," not "7.4× EBITDA."
What these cases demonstrate is narrower and more interesting: the four questions generalise, because they were never really about code. How hard is the problem to see? What does it truly cost to fix? How long does the wrong choice weigh on you? How attractive does the mistake look while you're making it? Any domain where decisions compound — finance, infrastructure, medicine, ops — can be read through those four questions. If someone calibrates anchors for another domain, the format is open and the license says go ahead.
Why hints, not names
Three reasons, in honesty order. First: the lessons are structural, and naming a company invites arguing about the company instead of the structure. Second: these situations are still live — litigation, restructuring, ratings in motion — and a scoring standard has no business piling on. Third: if the case only works because the name is famous, it isn't a case, it's gossip. Each of these works without the name. That's the test.
The four axes, the format, the three cases above, and the rules for domain branches — written up as a citable whitepaper. Free, like everything else here.
↓ Download the whitepaper (PDF)