The Probability Ledger

A shipment of steel sits in a bonded warehouse while two companies on opposite sides of an ocean argue over who pays a tariff nobody anticipated when the contract was signed. The supplier invokes a force-majeure clause; the buyer cites a price-adjustment provision; counsel for both sides reaches for the same instinctive question their predecessors asked a generation ago, what are our chances? What has changed is the answer. Where a senior litigator once offered a gut estimate dressed in qualifiers, a growing number of legal teams now produce a probability curve: a simulated spread of outcomes, each with a likelihood and a dollar value attached. The supply-chain dispute, long the messiest corner of commercial law, is being rebuilt around math.

This shift matters because supply-chain disputes are no longer rare or small. The International Chamber of Commerce registered 841 new arbitration cases in 2024, and the aggregate amount in dispute across its pending caseload reached an all-time record of US$354 billion, with construction, engineering and energy matters, the connective tissue of global trade, accounting for 44% of new filings (ICC). When that much value is in motion and 69% of the caseload is cross-border (Aceris Law), the gap between a good prediction and a bad one becomes a balance-sheet event.

The headline numbers tell two stories at once: disputes are getting larger and more international, and the tools to forecast them are going mainstream. Two-thirds of legal professionals reported using legal analytics in 2024, and every single user surveyed called it valuable (Lex Machina Legal Analytics Survey). To understand why simulation has captured the supply-chain bar specifically, it helps to remember how the work was done before.

The Old Way: Judgment in a Vacuum

For most of the twentieth century, evaluating a commercial dispute was an exercise in experienced guesswork. A lawyer read the contract, weighed the facts, recalled how a similar matter resolved years earlier, and produced a recommendation: settle, fight, or wait. The reasoning was often sound, but it was opaque, unrepeatable, and almost impossible to audit. Two equally senior partners could look at the same file and reach opposite conclusions, each defensible, neither falsifiable.

Supply-chain matters made this worse. A single cross-border contract dispute might hinge on the governing law of one country, the enforcement regime of another, a force-majeure clause whose meaning had never been tested, and currency and tariff variables that shifted weekly. The force-majeure provision, a clause "allocating the risk of loss if performance becomes impossible or impracticable" as a result of an unforeseen event, was rarely litigated until a crisis forced the question (Association of Corporate Counsel). Lawyers could describe the risks eloquently; they could not quantify them in a way a board or an insurer would trust.

A quieter discipline had been waiting in the wings. Decision-tree analysis, mapping each possible event in a case as a branch, assigning a probability and a value, and computing a probability-weighted "expected value", had long been popular in business but only slowly migrated to law (University of Cincinnati Law Review). One formalized version, Litigation Risk Analysis, was used by in-house departments to expose the key areas of uncertainty in high-exposure matters and to translate legal judgment into numbers (Litigation Risk Analysis). The expected value of such a tree is, in effect, the rational settlement value of a case assuming both parties are risk-neutral (American Bankruptcy Institute Journal). The method was powerful but laborious, and it captured only a handful of branches by hand.

The Shift: From a Single Estimate to a Distribution

The leap from a hand-drawn decision tree to a simulated outcome distribution is the heart of the present transformation. Monte Carlo methods, running a model thousands of times while randomly sampling each uncertain variable, let a legal team move from "we think we win about 60% of the time" to a curve showing the probability of every recovery between zero and the full claim. For a supply-chain dispute layered with tariff exposure, currency swings, and contested liability, that richer picture is the difference between negotiating blind and negotiating with a map.

Adoption has accelerated sharply. Roughly seven in ten law-firm professionals now report using some form of legal analytics, with usage nearly doubling since 2018 (Burford Capital). Crucially, the driver is client pressure, not novelty: 80% of users say analytics is required or expected by their clients, and 70% point directly to better litigation outcomes as the reason they adopted (Lex Machina Legal Analytics Survey).

The supply-chain sector is not an abstract beneficiary of this trend, it is one of its busiest customers. More than half of automotive businesses now name supply-chain issues as their single biggest source of litigation exposure: 61% flagged supply-chain litigation as a concern, driven by fights over pricing adjustments, tariff responsibility, and tariff-related cost recovery, while 52% named customs and tariff enforcement as their top regulatory risk (Global Legal Post, on the Dykema 2026 Automotive Trends Report). These are precisely the disputes where a probability distribution beats a hunch, because the underlying variables, duty rates, exemption odds, enforcement outcomes, are themselves uncertain and interacting.

The financial stakes of getting it wrong are not theoretical. Direct procurement disruptions cost organizations an average of US$16 million a year, and modeling by leading research firms estimates that disruptions erase roughly 45% of one year's EBITDA over the course of a decade for the average company (Conexiom, summarizing Coupa and McKinsey research). A single prolonged production shock can wipe out between 30% and 50% of a year's EBITDA in most industries (McKinsey & Company). When the downside is that large, the value of a tool that prices tail risk before it materializes is obvious.

What It Looks Like Now: Three Decisions, Modeled

Stripped of vendor branding, today's outcome-simulation workflows tend to support three recurring supply-chain decisions. Each takes a question that used to be answered by feel and reframes it as a distribution.

1. Cross-border dispute outcome prediction

Before committing to arbitration in a foreign seat, counsel can now model the realistic spread of outcomes given the governing law, the chosen institution, and the historical behavior of comparable matters. The data backbone is substantial: institutional caseloads, average proceeding durations of around 26 months for matters ending in a final award (Herbert Smith Freehills Kramer), and the distribution of claim sizes all feed the model. The output is not a verdict but a probability-weighted view of recovery, timing, and cost that a general counsel can defend to the board.

2. Contract-risk and penalty modeling

Penalty clauses, liquidated damages, and indemnities carry uncertain triggers. Simulation lets a team vary the probability that each clause fires, the magnitude of the resulting liability, and the correlation between them, then read off the chance that combined exposure breaches a threshold. This is the decision-tree expected-value logic of the 1990s, scaled to thousands of iterations and dozens of interacting variables.

3. Sanctions and tariff scenario comparison

This is the fastest-growing use, and the timing is no accident. Independent modeling of the 2025 tariff measures projects that the new duties would reduce the U.S. growth rate by 0.23 percentage points in 2025 and 0.62 percentage points in 2026 if left in place (Peterson Institute for International Economics). Broader trade modeling finds global trade flows contracting by between 5.5% and 8.5% depending on the scenario, with U.S., China trade falling by as much as 90% under a full-retaliation case (CEPR). Market researchers, meanwhile, put the odds of a global recession at around 40% (J.P. Morgan Global Research). Each of those figures is itself a scenario with a probability, exactly the kind of input a Monte Carlo engine consumes to compare, say, the cost of absorbing a tariff versus litigating who bears it.

The supply-chain response to these pressures is itself measurable, which gives simulations real-world anchors. In a 2025 survey of supply-chain executives, 45% of those facing tariff impacts said they were increasing inventories as mitigation, and roughly 30% were pursuing tariff-specific responses such as renegotiating with suppliers or seeking exemptions (McKinsey & Company). Those behavioral probabilities, how often a counterparty renegotiates rather than litigates, are precisely what an outcome model needs.

The Next Few Years: Faster, Cheaper, and Riskier to Trust

The trajectory is clear in the survey data. More than 95% of legal professionals expect generative AI to become central to their workflow within five years (LawSites, on Thomson Reuters research), and the same body of research shows AI tools already saving lawyers close to 240 hours a year on routine work (Thomson Reuters). As that capacity compounds, building and re-running an outcome simulation for a mid-sized supply-chain dispute will move from a specialist exercise to a routine first step, early case assessment by default.

Expect three developments. First, integration: 65% of analytics users already want to merge their analytics with other organizational data (Lex Machina Legal Analytics Survey), pointing toward live models fed by a company's own contract, shipment, and customs data. Second, scenario libraries: reusable tariff and sanctions templates that update as trade policy shifts. Third, the spread of simulation from disputes into deal design, modeling penalty exposure before a contract is signed rather than after it breaks.

That last figure is the crucial caution. More than 95% of legal and tax professionals say that letting AI represent clients or make final decisions on complex legal matters "would be a step too far" (Thomson Reuters Future of Professionals Report), and accuracy has become the top barrier preventing firms from increasing AI investment (Thomson Reuters). The reservation is well founded. Even the strongest prediction systems hover in the 80 to 90% accuracy range for narrow questions like whether a single motion will be granted, leaving a meaningful margin of error, and seasoned observers can already predict many outcomes 60 to 70% of the time from experience alone (Accumulated).

This is the genuine risk of the simulation era. Cross-border supply-chain disputes are often novel: a tariff regime that did not exist last year, a sanctions list rewritten last month, a force-majeure clause never litigated. In those settings the historical data that trains a model may be thin, and the model's confidence can outrun its competence. The discipline that mattered in the decision-tree era, interrogating every probability and treating the output as an argument rather than an oracle, matters more, not less, when the tool can generate ten thousand scenarios in a second.

Conclusion: The Number Is a Beginning, Not a Verdict

Outcome simulation has done something genuinely new for supply-chain legal strategy: it has made uncertainty legible. A general counsel can now walk into a board meeting with a distribution instead of an adjective, and a negotiating team can anchor a settlement to a defensible expected value rather than a feeling. With record dispute volumes, cross-border complexity, and a tariff environment that rewrites itself by the quarter, that capability is no longer a luxury.

But the most sophisticated users have learned the same lesson the early decision-tree analysts learned thirty years ago, only with higher stakes: the value is in the reasoning the model forces, not the certainty it appears to offer. The probability ledger is a powerful instrument for thinking. It is a dangerous substitute for it. The firms that thrive in the next few years will be the ones that run the simulation, study the tail, and still send a human to read the room.

Sources

1. International Chamber of Commerce, 2024 ICC Arbitration and ADR Preliminary Statistics
2. Aceris Law, Key Takeaways from 2024 LCIA and ICC Arbitration Statistics
3. Herbert Smith Freehills Kramer, ICC 2024 Dispute Resolution Report: Caseloads, Complexity, and Global Reach
4. Lex Machina, 2024 Legal Analytics Survey
5. Burford Capital, Growing Law Firm Adoption of Legal Analytics Boosts Certainty
6. Global Legal Post, Supply-Chain Disputes Top Automotive Litigation Risks in 2026 (Dykema study)
7. University of Cincinnati Law Review, Using Decision Trees As Tools for Settlement
8. Litigation Risk Analysis, Evaluating Legal Risks and Costs with Decision Tree Analysis
9. American Bankruptcy Institute Journal, Applying Decision Tree Analysis to Expedite Settlements
10. Association of Corporate Counsel, The Force of Force Majeure in Supply-Chain Agreements
11. McKinsey & Company, Risk, Resilience, and Rebalancing in Global Value Chains
12. McKinsey & Company, Supply Chain Risk Pulse 2025: Tariffs Reshuffle Global Trade Priorities
13. Conexiom, The Cost of Supply Chain Disruptions (summarizing Coupa and McKinsey research)
14. Peterson Institute for International Economics, The Global Economic Effects of Trump's 2025 Tariffs
15. CEPR, Roaring Tariffs: The Global Impact of the 2025 US Trade War
16. J.P. Morgan Global Research, US Tariffs: What's the Impact?
17. LawSites, Thomson Reuters Survey: Over 95% Expect Gen AI Central to Workflow Within Five Years
18. Thomson Reuters, How AI Is Transforming the Legal Profession
19. Thomson Reuters, Future of Professionals Report 2024
20. Thomson Reuters, 2025 AI Adoption Trends: What Changed?
21. Accumulated, Litigation Forecasting: Can Analytics Predict Case Outcomes?