Loading...
Loading...
Everything we show is derived from public SEC filings. Here is exactly how we collect, extract, and analyze the data — and where we fall short.
All compensation and governance data comes from official filings on SEC EDGAR. We do not scrape news articles, press releases, or third-party databases.
Stock prices for Total Shareholder Return (TSR) calculations come from Yahoo Finance. Market capitalization is sourced from Yahoo Finance and cached for up to 24 hours.
We use a deterministic-first approach. Structured HTML tables in SEC filings follow predictable patterns that rule-based parsers handle reliably. AI is used only as a repair layer when deterministic extraction is incomplete.
Filings are fetched directly from EDGAR using CIK + accession number lookups. We respect the SEC’s 10-request-per-second rate limit and identify ourselves with a generic bot user agent.
Custom parsers extract structured data from HTML tables in proxy filings:
When deterministic parsers fail to extract a table — typically due to unusual formatting, merged cells, or non-standard iXBRL markup — we fall back to an LLM. The AI receives the raw HTML and is asked to extract the same structured fields. Its output goes through the same validation pipeline as deterministic results.
Transparency note: Across our dataset, approximately 80% of extractions are fully deterministic. The remaining 20% use AI repair. We are actively improving parser coverage to reduce AI dependence.
Every extraction — whether deterministic or AI-assisted — passes through post-ingestion checks:
Our pay-for-performance verdicts (Aligned, Misaligned, Mixed Signals) analyze the relationship between executive compensation and shareholder returns.
Research basis: Our thresholds draw on Pay Governance peer alignment frameworks and ISS Relative Degree of Alignment (RDA) standards. For academic background, see the Harvard Law Forum on pay-for-performance alignment. We break down these frameworks and apply them to our full dataset in Are CEOs Paid for Performance?
Data freshness varies by filing type:
We believe in stating our limitations before anyone else does:
For full legal terms, see our Terms of Service. For definitions of compensation terms, see the Glossary.