How we build trustworthy financial data

The dataset spans the full SEC EDGAR universe of XBRL filers from 1993–present: every active company and every company that has since delisted, gone bankrupt, or been acquired. It is organized as two datasets — a 111M+-fact fundamentals core and a 78M+-row smart-money dataset — across 17 Parquet tables.

Filing forms covered

The fundamentals core is sourced from the SEC's quarterly EDGAR Financial Statements Data Sets plus the per-filing XBRL submissions — the same primary source the Commission publishes. Nothing is scraped from third-party aggregators. See the full dataset page for the pipeline and per-table schema.

We map 11,966 raw XBRL tags to 292 canonical concepts. Definitions are versioned in a taxonomy_guide table that ships with every Parquet bucket — so you can audit every transformation we apply, and unmapped tags fall through to a labelled Other rather than being silently dropped.

Worked example: Revenue

All five resolve to standard_concept = 'TotalRevenue'. Every fact also keeps its raw concept column — the exact XBRL tag the company filed — so you can always trace a standardized value back to source. This 5-row illustration is the level of detail we publish; the canonical name and definition of every concept lives in the data catalog.

When a company files a 10-K/A, the SEC treats it as a new filing — not an overwrite. Most data vendors collapse the amendment over the original, destroying the historical view. We keep both.

-- Apple FY2018 net income, original filing
ticker     fiscal_year  standard_concept  numeric_value  accepted_at
AAPL       2018         NetIncome         59531000000    2018-11-05T18:23:00Z

-- Apple FY2018 net income, after restatement (hypothetical)
ticker     fiscal_year  standard_concept  numeric_value  accepted_at
AAPL       2018         NetIncome         59300000000    2019-02-12T14:51:00Z

A backtest that ran on 2018-12-01 sees the first row only — the original $59.531B. A current dashboard sees the latest accepted value. Both are correct; both are queryable. The PIT discipline is what guarantees you get the right one.

“What was in the S&P 500 on March 1, 2014?” is a survivorship trap in disguise: screen the index by its current members and you have quietly excluded everyone who was dropped. Index membership is therefore tracked the same way facts are — historically, with effective and removal dates.

The index_membership table records membership spells for S&P 500 and Russell 1000 / 2000 / 3000 with an effective_date and a removal_date per spell, using half-open [effective, removal) interval semantics. A company that left and rejoined gets two spells, not a merged one — so a point-in-time universe on any date reconstructs the index exactly as it stood.

SELECT r.symbol, r.name
FROM index_membership im
JOIN references r ON r.cik = im.cik
WHERE im.index_name = 'SP500'
  AND '2014-03-01' >= im.effective_date
  AND ('2014-03-01' < im.removal_date OR im.removal_date IS NULL);

There is no is_sp500 flag — a single boolean can only describe one index at one moment, which is precisely the snapshot bias we avoid. Membership is always a JOIN on cik.

Foreign private issuers don't file 10-Ks. They file 20-F (and Canadian issuers file 40-F), often under IFRS rather than US GAAP, with their own concept names. Those filings flow through the same standardization pipeline: concepts map into the same canonical standard_concept vocabulary, and an is_foreign flag on the entity lets you include or isolate them. The result is that a US filer and a foreign issuer answer the same query the same way.

Every table is a column-oriented Parquet file with ZSTD compression — built for DuckDB, Polars, and Spark. A manifest.json ships alongside the data with the snapshot date, the last_updated timestamp, and a row count for every table, so any integration can detect fresh data automatically and verify it received the whole dataset.

The fundamentals core refreshes on the SEC's quarterly EDGAR cadence with amendments processed continuously. On the Institutional tier, filings carry an intraday accepted_at — acceptance timestamps at the moment the SEC published, not a date-only floor.

All four read from the same standardized core — and a single Stripe-issued token unlocks every one of them at your tier. There is no per-channel divergence in the numbers, because there is only one set of numbers.