What the platform owns, stated the way we would want to diligence it.
Every figure below is governed by a single verification protocol — counts change only when the underlying registry or corpus changes, and each carries its definition next to its number. Detailed reconciliation is available in diligence.
Jurisdiction-specific outcome, duration, and settlement models — each one narrow by design (one jurisdiction × one case type × one target) and individually cleared through AUC ≥ 0.70 temporal-holdout promotion gates. The registry publishes its own denominator: entries that failed the gate stay recorded as experimental, stub, or failed.
Real filed court and regulatory records, deduplicated on composite source keys with provenance carried through the training pipeline. Exact figures are disclosed in diligence with methodology attached; public surfaces describe scale qualitatively by policy.
Judicial assignment as a first-class model feature across all 50 US states and all federal circuits — the venue-level variance that pooled national statistics average away.
United States, Canada, United Kingdom, and Australia in production, with additional markets in data-ingestion stages — the common-law footprint where institutional legal capital concentrates.
Filings covering the outcome-prediction and data-infrastructure methods underlying the platform.
Criterica Intelligence is developed and owned by Splitifi. The production model fleet, the court-record corpus, and the platform infrastructure are Splitifi-built assets; Criterica is the institutional market identity under which they are organized. No third-party license sits between the platform and its IP.
Model and registry definitions, gate methodology, and the public correction history are documented on the Intelligence site's registry and methodology pages.