Twenty-eight worked cases across eight domains.

Statins & Hospitalisation → The correct average applied to the wrong individual.

A policy that works on average is applied uniformly — including to patients for whom it is harmful, neutral, or unnecessary. Population statistics used to make individual decisions.

Iatrogenic Medications → The statin isn’t fighting the disease. In many cases, it’s fighting another drug.

An audit attributed a large share of adverse events to underlying conditions. The causal model found they were drug-drug interactions in the prescribing cascade. The reserve was overstated; the clinical liability was understated.

Adverse Event Attribution → The NSAID raises AKI risk in the population. Did it cause this patient’s AKI?

A pharmacovigilance signal is a population-level association. The clinical and legal question is whether this exposure caused this harm. Two different questions, two different formal structures.

Sepsis Dynamic Treatment Regimes → The vasopressor was given to the patients who needed it — so observational data shows it kills.

Treatment confounded by indication is the classic Rung 1 trap. The covariates that selected for treatment also selected for outcome; standard adjustment cannot reach the counterfactual. A dynamic regime model can.

Immunotherapy vs Chemotherapy → The biomarkers that select the patient also select the response.

Selection effects make immunotherapy look better than it is in head-to-head observational comparisons. The model separates selection from effect — the patients who would have responded to chemo, do.

Drug Repurposing — Transportability → The trial population is not the target population.

A drug’s effect estimated in one population does not transport unaltered to another. The transportability formula closes the gap when the causal structure is known and the differences are characterized — and is silent on what to do when they aren’t.

Treatment-Resistant Depression → After two failed antidepressants, “what works on average” is no longer the right question.

The patient has demonstrated what doesn’t work for them. The next decision is a Rung 3 counterfactual: which option, on this individual, given their non-response pattern. Population statistics applied to a non-representative individual.

Property Insurance → The intervention that makes things worse.

A board considering a 15% rate increase on a $412M coastal portfolio. The causal model showed the increase would accelerate adverse selection — and identified the cheaper fix the register couldn’t see.

Insurance Reserving → Chain-ladder tells you the pattern. It cannot tell you why the pattern is changing.

Social inflation, attorney involvement, severity creep, or a mix — each carries different reserve implications. Regulators are increasingly asking for the why; the standard tooling cannot supply it.

Insurance Attribution → The expert who was right for the wrong reason.

The senior adjuster apportions every claim correctly — drawing on jurisdiction, injury pattern, and two decades of intuition. When she retires, the reasoning goes with her. The correlate may not hold in the next case.

Self-Insurance Decision → Most self-insurance analysis treats future losses as a fixed distribution.

The act of self-insuring changes the distribution — through claims behavior, loss-prevention investment, and risk culture. A causal model represents the change; an actuarial mean does not.

Bank Churn → The program that looked like it worked.

£2.4M spent on a retention campaign. The data showed it working. The causal model asked the right question: would those customers have stayed regardless?

Credit Risk → We measured program effectiveness by looking at enrolled borrowers.

Borrowers selected into the program based on intent to stay. The observed effect is mostly the selection, not the program. The standard analysis confounded the two; the causal model separates them.

M&A Due Diligence → The team that resisted the retention clause is the most capable team.

The negotiating signal is informative about exactly the variable the acquirer wants to retain. A causal model represents how the diligence process itself selects on the outcome it is trying to measure.

Rent vs Buy → A rate rise makes buying more expensive. It also cools prices.

The two effects move in opposite directions for the buyer. A spreadsheet that treats rates and prices as independent inputs answers a different question than the one the decision requires.

Utility Wildfire Risk → The utility knew wildfire was a top risk. It had a score. It had a regulatory framework.

None of it connected an equipment-deferral decision to a fire-probability change. The risk register described what was happening; it did not describe what to do, and it could not be queried that way.

Utility Grid Risk → Capital deferral decisions are made as if the grid is static.

Load is shifting, mix is shifting, and the controls available now will look different in five years. A static reliability model assumes the deferred asset will be needed for the same job; the causal model represents what happens when the job changes.

Asset Reliability → The control that prevented the problem you can’t see.

Three unplanned outages looks like a failure. The causal model asked how many would have occurred without the program. The KPI was wrong because the intervention was working.

NIST CSF 2.0 → Investment in the wrong priority.

$4.2M to allocate across two NIST functions. The maturity assessment said both needed investment. The causal model said one reduced breach probability by 61%, the other by 11%.

GDPR → Compliance exceptions are handed out like chiclets. Most don’t survive a regulator’s causal question.

An exception that passes the legal test is not the same as an exception that survives a regulator asking did this practice cause this harm. The causal model is what supports the answer in a form a regulator accepts.

CCPA / CPRA → The compliance checklist passed. The opt-out mechanism was broken.

A checklist verifies that controls exist. A causal model asks whether they work — whether the opt-out causes the requested outcome, or whether something downstream undoes the opt-out silently. Two different questions.

Causal Evidence → The regulator asks what caused the harm. Association is not the answer.

Regulators in financial services, healthcare, environmental liability, and employment law increasingly distinguish between statistical association and causal attribution. A structural causal model is the artifact that supplies the answer in defensible form.

Supply Chain Risk → The risk you didn’t know was there.

Three suppliers in three countries. The scorecard called it diversification. The causal model found one fabricator with three labels — $23M in concentration risk scored as managed.

Quality & Defect Attribution → The defect rate is 2.3%. Which line, which material, which operator?

A predictive model can tell you defects are correlated with night shift on line 4. A causal model tells you whether changing the line, the shift, or the supplier reduces the rate — and by how much.

Training Effectiveness → The training completion rate is 94%. The performance metric was unmeasured.

Completion was answerable from existing systems; the actual reason the program existed was not. The cheaper metric won by default. A causal model is what isolates the training effect from everything else that affects performance.

Climate & ESG Risk → Climate risk has a causal structure. The risk matrix does not represent it.

Climate risks share drivers, interact through mechanisms, and fire jointly in the tail. The standard register scores them as independent rows. The causal model represents the interactions the register erases.

Criminal Causation → The expert’s statistics are accurate. Their logic is backward.

A frequency stated forward (the probability of the evidence given innocence) is not what the question asks (the probability of innocence given the evidence). The prosecutor’s fallacy is a Rung-error; the causal model is what makes the direction explicit.

Elective Sequencing & Mastery → The elective that looks most effective is the one chosen by students most likely to succeed regardless.

Self-selection makes any observational comparison of elective effectiveness biased in the direction of the most-selected option. The causal model isolates the elective’s contribution from the student’s prior trajectory.