Research Methodology

Analysis begins with primary documents. Interpretation follows after facts, periods, and calculations have been reconciled. Published research is drafted, reviewed, and approved by people. Software does not independently determine the thesis.

Financial data methodology

Issuer identity and filing history come from SEC submissions. Financial facts come from SEC companyfacts XBRL. Requests are made server-side with an identifiable User-Agent, throttled below the SEC’s 10 requests/second maximum, with retry on 429/503. Identical filings are not re-downloaded when ETags allow.

Income-statement metrics use duration facts (quarter or annual) from the latest matched period. Balance-sheet metrics use instant facts aligned to the latest assets date. Annual facts are not compared with quarterly facts. Amended forms (10-K/A, 10-Q/A, 8-K/A) are stored; current views prefer the latest valid fact for a period.

Free cash flow

Default Belvedere FCF = net cash provided by operating activities minus payments to acquire property, plant and equipment. The figure is labeled “Belvedere calculation.” It is not company-defined non-GAAP FCF. FCF margin is Belvedere FCF divided by revenue for a matched period. Trailing-twelve-month figures are used in rankings only when four comparable quarters exist.

Valuation methodology

Enterprise value, EV/Revenue, EV/EBITDA, P/E, and FCF yield require a licensed market-price source. SEC EDGAR does not provide live prices. Until such a provider is configured, valuation multiples are omitted rather than estimated. Diluted shares, cash, and identified debt remain filing-based inputs, not a completed valuation.

Research Profile

The Research Profile Index is an equal-weighted average of Financial Disclosure, Operating Visibility, KPI Transparency, Investor Information, and Financial History, each 0–100. Bands are Apex, Established, Developing, Limited, and Insufficient. The index measures researchability of public information, not investment attractiveness. See Research Profile.

Ranking methodology

Each live ranking publishes universe, eligibility, formula, exclusions, and data date. Rankings cannot be purchased. Sponsored research clients cannot buy placement or score changes. See individual ranking pages for definitions, including the Belvedere Undercoverage Screen (a proxy, not an analyst census).

Research Radar

Radar items are arithmetic comparisons of sourced metrics across comparable periods. A signal is omitted when a prior comparable period does not exist. Publication on the Radar is not a thesis, rating, or recommendation. See Research Radar.

Research process

Initiation establishes a monitoring frame from primary documents. Updates follow new filings. Human review is required before publication of conclusions. Data refresh occurs when the SEC extract is rebuilt.

Evidence and sources

Important figures retain concept, taxonomy, unit, period, form, filed date, accession number, and a link to the SEC filing index. Calculated figures retain a formula. Secondary press is not treated as a substitute for a filing fact.

AI / technology

Machine-learning tools may assist extraction, normalization, calculations, filing monitoring, and quality control. They can err. Outputs are inputs to a controlled workflow, not authoritative sources. See the AI policy.

Research independence

Belvedere may receive compensation for sponsored research. Compensation does not purchase ranking position, favorable conclusions, valuation outcomes, or editorial control. Issuers may provide facts and identify errors; they may not edit analytical conclusions. Final regulatory language is subject to legal review.

Corrections

Material errors are dated and explained. See the corrections policy.