I audited 463 SEC filers to figure out why nobody parses geographic revenue correctly
Description and use
The edgar‑geo‑revenue library automatically extracts geographic revenue data from SEC 10‑K filings, overcoming XBRL gaps, unpredictable table locations, and inconsistent labels.
Problem
Programmatic extraction of geographic revenue from SEC 10‑K filings is unreliable because filings lack standardized location, table shape, or label vocabulary.
Outcome
The library extracts geographic revenue for 262 of 463 SEC filers, handling XBRL gaps, variable table locations, and diverse label vocabularies, and outputs clean JSON for downstream analysis.
Workflow
- Set a valid SEC User‑Agent with contact info via egr.set_user_agent("Name
<email>") - Call egr.fetch_geo_revenue("TICKER") to download and parse the filing
- Receive a JSON object containing fiscal_year, total_revenue_usd, segments, region_segments, and pct_coverage
- Validate results against the filing’s reported revenue for sanity checks
Evidence
Run against 463 filers: 262 had a usable disclosure. The rest largely don’t disclose one at all — a legitimate outcome, not a parsing failure.
XBRL tagging is often incomplete. Many filers disclose the split in rendered statements but tag little in XBRL
- Record date: 2026-09-08T01:23:01.873Z