A number is only useful if you can see how it was made.

This page describes the sources behind every report, the exact rules that turn those sources into a 0 to 100 risk index, and the limits of the whole thing. It is written so that a reader can disagree with a score and know precisely which input they are disagreeing with.

Data Sources

SEC EDGAR. The primary source. Company submissions, financial statement data (XBRL company facts), annual and quarterly reports, current reports on Form 8-K, registration statements, and insider Forms 3, 4 and 5.
US Treasury OFAC. The sanctions list, screened against company names.
Public registry and naming records. RDAP lookups, archived WHOIS history, and certificate transparency logs for the domains attached to an issuer.
GLEIF. The Global Legal Entity Identifier registry, for legal identity and corporate relationships.
Market data. A third-party provider supplies last close, confirmed splits, public float and short interest. Price data is used for context only and never feeds the risk score.

The Risk Index

The score is the sum of six independent components, each computed from the filings with fixed thresholds. Same inputs, same score, for everyone. It is a screening heuristic, not a forecast.

Red flags

Weighted counts of danger phrases in the last 12 months of filings (going concern, ATM, reverse split, delisting), each capped.

Dilution

Trailing 12-month change in shares outstanding: +50% scores 20, +25% scores 14, +10% scores 8, +3% scores 4.

Runway

Months of cash at the current burn: under 6 scores 14, under 12 scores 8, under 24 scores 3.

Cash drain

A 30% or larger year-over-year fall in cash on hand scores 5.

Chronic pattern

Flags that have appeared for years rather than this cycle add 2 each, capped at 6.

Official notices

Critical-item 8-Ks in the trailing 12 months (bankruptcy, delisting, non-reliance) score up to 15.

Component totals are summed and capped at 100. Bands: quiet 0 to 24, guarded 25 to 44, elevated 45 to 64, severe 65 to 100.

How Red Flags Are Counted

A phrase scan alone produces false positives, because words like "going concern" appear in boilerplate and in descriptions of other companies. Where the documents are available, the report opens the linked filing and filters obvious boilerplate before counting, and shows both the verified count and the raw full-text count so the difference is visible. When verification is not available, the raw count is used and labeled as such.

Update Cadence

Filings are fetched from EDGAR as they are requested and cached for up to 24 hours. Derived reports (the delta, the verdict) are cached for about six hours. When a company files something new, the cache for that company refreshes on the next request. Market data slots refresh on their own schedule, governed by the provider's rate limits, and the report tells you when a slot is waiting on a quota rather than inventing a number.

Limits

It reads documents, not intentions. The score describes what a company reported and what it has done. It says nothing about where the price goes.
It can be incomplete. Some issuers file irregularly or in formats that are harder to parse. Missing data is shown as unavailable, never as a clean result.
It is not advice. Nothing here is a recommendation to buy, sell or hold any security. Scores are a starting point for your own work against the cited filings.
It can be wrong. A parsing error or a stale cache can produce a bad number. When you find one, we fix it.

Corrections

If a report misstates something from a filing, email contact@decoded.report with the ticker and the specific line. Corrections are made against the source document. See also About for the editorial standards behind the tool.

DETERMINISTIC SCREENING OVER PUBLIC FILINGS. NOT INVESTMENT ADVICE. NOT A FORECAST OF PRICE OR PERFORMANCE.