Methodology

How Repleader collects records, groups cases into matters, decides what to publish, and labels what it computes.

Sources

• Court records: docket metadata, docket entries, parties and opinions from Free Law Project CourtListener. In scope are federal district courts and the Judicial Panel on Multidistrict Litigation, for civil nature-of-suit codes covering product liability, personal injury, securities, antitrust, consumer and related cases.
• Enforcement: releases from the SEC, FTC, Department of Justice, CFPB and the attorneys general of ten large states. Federal releases are public domain, so we store their text. State releases appear as headline, summary and link.
• Company filings: SEC EDGAR submissions and financial data, including litigation reserves reported in filings.
• News: headline, publisher, date and link only, from publisher RSS feeds that permit it.

Sourced and modeled values

Every value on a page is one of two kinds. A sourced value is copied from a record and carries a link to it. A modeled value is computed by Repleader, is labeled modeled, and names the method version that produced it, such as inertia-1.0.

Matters

A matter is a group of related cases. Cases join the same matter in one of these ways:

1. They share an MDL number, from the docket or from the text of its entries.
2. They name the same defendant, matched exactly, with the same nature-of-suit code, in at least 3 dockets across at least 2 courts, each active within 90 days.
3. They are securities cases against the same issuer filed within 120 days of each other.
4. An editor merged or split them by hand, recorded in a reviewed list.

Matter titles are the verbatim MDL caption, or the defendant name and nature-of-suit label. Titles are never written by software.

Matching companies

A case links to a company only by an exact normalized name, an SEC CIK number, or a reviewed alias. Repleader never merges two names because they look similar.

What gets published

A page is published only when it meets a publishing floor: enough distinct sourced facts, a data table, a dated timeline with sources, several primary identifiers such as docket numbers, and links to primary sources. Pages below the floor are not published, are left out of sitemaps and feeds, and are removed if they later fall below it.

Summaries

Summaries are drafted by software and checked by automated rules against the linked source. A summary that fails any check is not shown. See principles.

Methods

• Inertia: the measured activity series for cases and matters, and the Repleader index derived from them (inertia-1.0, backtest pending).