Contract risk analysis

Clause categories used

A subset of CUAD's actual 41 categories carries the contract-risk story. Weights are stored as data, not hard-coded.

Rule-based scoring

Each rule evaluates an annotation (presence, absence, or a keyword signal) and contributes points. Scores roll up per contract and per matter.

CREATE TABLE core.ref_risk_rule (
  rule_id     STRING,
  clause_category STRING,
  signal      STRING,   -- PRESENT | ABSENT | KEYWORD
  pattern     STRING,
  points      INT,
  rationale   STRING
);

-- Examples
-- ('R1','Limitation of liability','ABSENT', NULL, 40,'No cap on liability')
-- ('R2','Indemnification','PRESENT', NULL, 25,'Indemnity present; review scope')
-- ('R3','Governing law','KEYWORD','foreign|arbitration', 10,'Offshore forum')

Risk bands: High ≥ 60 Medium 30–59 Low < 30.

Analytics views

CREATE OR REPLACE VIEW marts.mart_matter_risk AS
SELECT m.matter_id, m.matter_name, c.contract_id, c.contract_type,
       SUM(CASE WHEN r.signal='ABSENT'  AND f.answer='No'  THEN r.points
                WHEN r.signal='PRESENT' AND f.answer='Yes' THEN r.points
                WHEN r.signal='KEYWORD' AND f.clause_text ILIKE '%'||r.pattern||'%'
                     THEN r.points ELSE 0 END) AS risk_score
FROM core.dim_matter m
JOIN core.bridge_matter_contract b USING (matter_id)
JOIN core.dim_contract c USING (contract_id)
JOIN core.fact_clause f ON f.contract_id = c.contract_id
JOIN core.ref_risk_rule r ON r.clause_category = f.clause_category
GROUP BY 1,2,3,4;

Because dim_matter and fact_clause carry matter_id, the same row access policy scopes these analytics — an attorney's risk report only covers their authorized matters.

Full SQL: sql/04_risk_views.sql.

Illustrative output

Values are illustrative; the demo generates them from the loaded contracts.