Damages quantum estimator

Commercial dispute teams need an exposure figure early, often before discovery closes and long before any settlement number is on the table. A damages quantum estimator produces a bounded financial range from the contract language that governs recovery, the breach facts already in the file, and outcomes from comparable matters. The output is built for quality review: every estimate names the contract term span it relied on and a comparable matter identifier. When terms or facts are too thin to support a defensible range, the estimator returns empty rather than a placeholder number that could be mistaken for counsel's reserve.

The estimator does not replace judgment on reserve, settlement strategy, or trial theory. It gives dispute counsel, finance, and risk committees a structured starting point that can be traced, challenged, and updated as the record develops.

Exposure conversations often begin with spreadsheet scenarios, partner memory of "similar" cases, or a single headline award from a news search. Those inputs rarely travel with the reasoning that produced them. Finance asks for a reserve; litigation responds with a wide band and little documentation of what drove the floor or ceiling.

The failure mode is not missing data alone. It is missing linkage between data and the legal theory of recovery. Contractual caps, limitation clauses, notice requirements, exclusive remedies, and measure-of-damages language can compress or expand quantum by orders of magnitude. Breach characterization (material vs. partial, anticipatory vs. actual, direct vs. consequential) changes which heads of damage are even arguable. Comparable outcomes only help when the matter is close on contract type, forum, industry, and remedy sought.

Without explicit citations to term spans and comparable IDs, an exposure range cannot survive scrutiny from auditors, insurers, or opposing counsel who will ask why this case is not the one where damages were capped at fees paid.

The workflow assumes three input layers, each with minimum completeness thresholds.

Contract terms. The system ingests the operative agreement (and amendments, side letters, and order forms where they modify liability). It extracts spans relevant to quantum: limitation of liability, liquidated damages, indemnity scope, warranty remedies, termination consequences, and any clause that defines or excludes categories of loss. Each span is stored with document reference and character offsets or clause numbering so downstream estimates can cite "Agreement §8.2(b), as amended by SOW #3."

Breach facts. Structured fields capture what happened, when, and with what contractual effect: performance shortfall, missed milestones, IP misuse, confidentiality breach, and similar categories aligned to the breach timeline reconstruction output where available. Facts that are allegation-only versus documented are flagged so the range can reflect confidence, not just magnitude.

Comparable dispute outcomes. The model queries outcome corpora and docket analytics for matters that match on contract family, claim type, forum, and remedy. Lex Machina and Westlaw supply aggregated award bands, time-to-resolution, and motion success patterns where the matter profile fits. When a matter-level match exists, the estimator records a comparable matter ID (docket number, Lex Machina matter key, or internal war-room ID). Weak matches are excluded rather than blended in silently.

The engine combines term constraints (hard ceilings, excluded damage categories), fact-driven heads of loss (direct, consequential where not excluded, restitution, specific performance value), and comparable distributions to emit a low, mid, and high exposure figure with explicit assumptions stated in plain language.

The outcome target is quality, not volume of numbers. Every non-empty estimate must include:

• Contract term span: the exact clause or span that caps, defines, or enables the damage category used in the calculation.
• Comparable matter ID: at least one cited comparable with enough metadata (forum, claim type, outcome class) for a reviewer to pull the underlying matter or summary.

If either layer is missing for a proposed head of damage, that head is dropped. If no head survives validation, the estimator returns empty. Empty is the correct result when, for example, limitation language is ambiguous pending construction briefing, breach dates conflict with notice periods, or no comparable within defined similarity thresholds exists.

Reviewers should treat empty output as a work order: obtain missing terms from contract evidence package assembly, tighten breach facts on the timeline, or widen (carefully) the comparable search via precedent and case law retrieval before asking again.

Estimates that pass validation still carry uncertainty bands. The mid figure is a working hypothesis, not a prediction of award or settlement.

No single product owns the full stack. Teams typically compose:

| Role | Typical sources | |------|-----------------| | Contract term extraction and obligation graph | Ironclad (repository and clause metadata), exported agreement PDFs with manual span confirmation | | Outcome and litigation analytics | Lex Machina (motion and damages patterns by judge, party, and claim type), Westlaw docket and verdict research | | Custom scenario logic | Excel models maintained by finance or litigation support (sensitivity tables, NPV of lost revenue, discount rates) | | Orchestration | Internal workflow that merges structured breach facts, validated term spans, and comparable IDs before any LLM or rules engine synthesizes the narrative range |

Ironclad helps when the dispute turns on what was actually signed across a chain of order forms. Lex Machina helps when forum and judge behavior materially affect recoverable categories or early disposition. Westlaw remains the fallback for bespoke comparables and emerging case law on measure of damages. Excel models remain valuable for damages theories that depend on financial assumptions (lost profits, market price, mitigation) that analytics platforms do not fully parameterize.

The estimator should read from these systems via stable identifiers, not pasted summaries. A Westlaw KeyCite or Lex Machina matter ID attached to an estimate is auditable six months later; a paraphrased "similar case in Texas" is not.

Litigation counsel and finance jointly own the reserve. The estimator informs that conversation; it does not close it.

A practical sequence:

1. Ingest contract package and breach timeline; reject run if mandatory fields are incomplete.
2. Generate range with term-span and comparable citations per head of damage.
3. Stress-test the theory using the legal position stress tester: alternate constructions of limitation clauses, mitigation arguments, and forum shifts that would move the band.
4. Document what counsel adopted, adjusted, or rejected for the reserve memo, preserving citations even when the reserve is set above or below the model mid-point.
5. Refresh when new facts, amended pleadings, or dispositive motion outcomes change term construction or comparability.

Finance teams should expect explicit language in reserve papers: "Model range $X–$Y based on §[span] and Comparable [ID]; reserve set at $Z due to [litigation strategy / insurance / settlement posture]." That sentence pattern keeps AI-assisted estimates inside governance norms for public companies and insured portfolios.

The estimator does not value punitive damages where unavailable, resolve choice-of-law conflicts, or model counterclaims without separate inputs. It does not produce privilege work product suitable for direct disclosure to opposing parties without counsel review.

Use it alongside breach chronology, evidence packaging, precedent retrieval, and position stress-testing so exposure is never a single isolated number. Together, those tools keep dispute preparation aligned: facts on a timeline, authorities on the record, contract language pinned to spans, and financial exposure tied to both.

When terms or facts are insufficient, prefer empty output and a clear list of gaps over a false precision that counsel must later unwind under audit or settlement pressure.

[REDACTED]