The full corpus, as a data feed
Language-stability scores for 4,930 US filers —
deterministic, corpus-normalized, auditable to the underlying filings. Built for quant/factor
research and credit monitoring.
What you get
Per-company year-over-year 10-K scores (level + history), the live control ceiling,
currently-flagged status (480 companies today), and filing metadata.
10-Q coverage is in progress.
Methodology discipline
Scores are computed from filings as published (point-in-time; no restatement lookahead),
normalized against the whole corpus, and reproducible — every number maps to a real
function in the scoring code.
Methodology.
Delivery
REST API (JSON), CSV export, or a custom feed (S3/SFTP drop, bespoke schema, SLA) for
institutional use.
API docs.
Validation, honestly stated: in a full-period backtest (2000–2026),
the most language-stable quintile earned
93.2 bps/month of alpha
against a five-factor adjustment (market, size, value, momentum, quality; t=9.16). Known limitation we disclose up front: the backtest
universe excludes delisted filers (
survivorship bias inflates the long side); correcting for it
concentrates the edge in small-caps. Full derivation, caveats, and reproduction steps:
the validation write-up and
methodology. Research data — not investment advice.
Two ways in
The Desk plan ($499/mo) includes bulk API access, full-corpus
sweep alerts, unlimited watchlist, and 5 seats. For a custom feed, schema, or license terms,
talk to us directly.
Want a taste first? Free sample CSV ·
live flagged list ·
worked examples