Basin
A consolidated E&P database of US oil & gas producers, built from SEC filings, with a citation behind every number.
Stack
- Python
- SEC EDGAR
- XBRL
- SQLite
- FastAPI
- Claude API
Build log
- 01
The panel: reserves, production, realized prices, per-barrel costs and capex for 91 filers across two cohorts, consolidated into one queryable table from 3,679 cited filings. The data is public and free already; it is unusable because it is scattered across hundreds of pages of prose, tables and inconsistently-tagged XBRL, restated and relabelled year to year.
- 02
Provenance: every one of 19,729 facts carries the accession number, form, fiscal period, page, line and verbatim source text it came from, and every current cell has been located in the document it cites — 97% found verbatim. A citation is a claim until it is checked, so the checking is a build step rather than a promise.
- 03
Comparability: filers are badged on two axes measured from their own filings rather than inferred from domicile — whether the facts layer can read them at all (66 us-gaap, 19 ifrs-full), and whether two numbers mean the same thing (81 SEC Subpart 1200, 8 Canadian NI 51-101). Reserves under forecast prices on a 2P headline are not reserves under a trailing 12-month average on a proved one, and a panel that silently mixes them is worse than no panel.