SEC EDGAR XBRL Financial Statements API

It is the difference between reading about a bank's finances and reading the bank's own numbers. The U.S. Securities and Exchange Commission requires every registered issuer to file its financial statements in XBRL - a tagged format where each number carries its concept, unit and period - and this dataset is those tags, delivered as structured records rather than PDFs. Four surfaces do the work:

• Company facts - every XBRL fact a registrant has ever reported, across all its taxonomies. JPMorgan Chase (CIK 19617) returns 918 us-gaap concepts plus dei, invest, srt, ffd and ecd, each with label, description and units.
• Single-concept histories - one tag's full timeline, such as Deposits or NetIncomeLoss, without dragging the whole file.
• Frames - a whole-market snapshot of one concept for one period; the verification probe of us-gaap Assets in USD for CY2025Q3I returned 5,628 data points, every filer that reported that instant.
• Filings index - each registrant's submission ledger with accession number, filing date, report date, form type and primary document, so facts resolve back to their source documents.

Because every fact keeps its fy, fp, form and filed stamps, statements rebuild as originally reported rather than restated - the version analysts actually need when a restatement muddies a trend line. Get a sample of this dataset cut to your tickers before anything else.

One fact per row, provenance attached. This is a real reading from the catalog research pass - JPMorgan Chase's total assets at the close of 2008:

cik       : 19617
entityName: JPMORGAN CHASE & CO
tag       : Assets
units     : USD
end       : 2008-12-31
val       : 2175052000000
fy        : 2009
fp        : Q2
form      : 10-Q
accn      : 0000950123-09-032832
filed     : 2009-08-10

$2,175,052,000,000 in total assets, tagged under us-gaap, stated in dollars, reported on the second-quarter 10-Q, traceable to accession 0000950123-09-032832 accepted on 2009-08-10. Nothing about that row required interpretation: the concept names what was measured, the unit says in what, and the form and accession say who said it and when. Pull the same tag across periods and the balance sheet becomes a time series; pull it across CIKs for one instant and you have the whole market's asset ranking in one frame. That double movement - deep in time for one issuer, wide across issuers at one moment - is what separates structured XBRL facts from every aggregator's cleaned-up summary.

Fourteen documented fields define a fact, each definition checked by calling the live endpoints during the August 2026 research pass rather than inferred from documentation. The design splits cleanly in two: five identifying columns (cik, entityName, taxonomy, tag, units) say whose number this is, while nine measurement and provenance columns (end, val, accn, fy, fp, form, filed, plus label and description) say what it is and where it came from.

Two consequences matter for pipelines. First, joins hold without a crosswalk: cik is stable for a registrant's lifetime, so a panel built on it survives renames and reorganizations. Second, point-in-time logic is native, not reconstructed - filter on form and filed and the statement a portfolio manager would have seen on filing day reassembles exactly.

Duration concepts (income-statement lines) additionally carry a start date alongside end, and some frames carry a frame label marking inclusion in the standardized period snapshots; both fold under additional fields on request.

Geography - every SEC-registered issuer: US-listed companies including foreign private issuers that file with the commission. For banking work that means the entire listed US diversified-bank universe, from the money-center giants down to the smallest regional still trading, plus custodians and trust banks filed under adjacent SIC codes.

Temporal - the full XBRL reporting history per company. The research pull verified JPMorgan Chase facts running back to 2008, and depth varies with each filer's own XBRL history rather than a fixed window. The filings index covers recent submissions; older documents remain reachable through EDGAR's daily indexes, so a decade-plus panel is assembled rather than assumed.

Granularity - one fact per company per concept per period, the finest cut financial disclosure allows. Frames add the cross-section: one call returns every filer reporting a given concept at a given instant, which is how peer sets get assembled without looping over tickers.

Set against the wider Datadory catalog - mean quality score 7.81 across all 1,744 datasets - this record scores 10/10, one of only 145 to reach the ceiling, carried by complete field documentation, live-verified pulls and coverage no vendor republishes with more authority than the filer itself. See best diversified banks datasets.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; the field dictionary above travels unchanged across all three. Structured payloads suit products that surface a bank metric inside an app, bulk files suit analysts loading full panels once and joining them to their own macro tables, and warehouse delivery suits teams running system-wide screens in SQL against filing-day freshness. Cadence is a settings conversation, not a re-integration project - tighten it around earnings season, relax it when nothing is landing.

• Fundamental research and valuation - line-item income statements, balance sheets and deposit trends rebuilt per bank from the filer's own tagged figures, not a vendor's recut of them.
• Point-in-time backtesting - form, fy, fp and filed stamps let a quant reconstruct exactly what was knowable on a given date, the discipline that separates honest backtests from look-ahead leakage.
• Peer benchmarking - frames return one concept for every filer at one instant, so asset, deposit or provision rankings across the whole listed universe arrive in one pull.
• Regulatory-adjacent monitoring - new filings surface as they are accepted, so competitor disclosures, segment changes and risk-factor shifts register on the day they land.
• Entity resolution - cik and entityName act as join keys onto FDIC registry rows, Federal Reserve series and proprietary credit files, collapsing name-matching projects into a one-column lookup.
• Citation-grade sourcing - every quoted figure resolves to an accession number, which is why journalists and academics prefer it over secondary charts.

Investors and quant researchers (relevance 3) read point-in-time GAAP statements and filings for fundamental research on listed banks; see investors quants use cases. Data scientists and ML engineers (relevance 3) pull full reporting histories for every listed issuer as model features; see data scientists use cases. Developers and data-product builders (relevance 3) wire documented JSON structures into data products; see developers builders use cases. Journalists, academics and students (relevance 3) source every quoted figure from official filings rather than aggregator sites; see journalists academics use cases. Market researchers and consultants (relevance 2) benchmark listed peers' results from standardized facts; see market researchers use cases. Competitive intelligence and product teams (relevance 2) track competitors' filings and disclosures as they land; see competitive intel product teams use cases.

Within diversified-banks data this record owns as-filed GAAP detail: one bank, one concept, one period, straight from the audited submission. The neighbors own different jobs. FDIC BankFind Suite API reports regulator-calculated quarterly financials for roughly 4,700 insured institutions reaching back toward the 1930s - including private and non-filing banks EDGAR never sees - but not the line-item statement detail a filing carries. FDIC Quarterly Banking Profile aggregates those institutions into industry totals. ECB Consolidated Banking Data contributes 66,139 euro-area series, and BIS Locational Banking Statistics maps cross-border exposures no national filing contains. None of them hand you a single bank's own tagged income statement the way a companyfacts pull does. The head-to-head with the FDIC product is worked through in FDIC BankFind Suite vs SEC EDGAR XBRL; the whole shelf sits on the diversified banks data hub.

Four honest caveats. First, listed issuers only: coverage runs to SEC registrants, so privately held community banks appear in regulator registries but not here - pair with the FDIC products when the question includes them. Second, frames report only what companies actually filed: a concept-period combination absent from the snapshot means nobody disclosed it that period, not that the data is missing. Third, some bank-specific concepts ride outside us-gaap: certain disclosures land under the ffd taxonomy whose coverage deserves a check before you depend on it. Fourth, depth follows each filer's XBRL history: a small regional bank's earliest tagged facts may postdate a giant's by years, so panels should be built per issuer rather than assumed uniform.

None of that changes the join keys, which stay stable across a registrant's lifetime. Name your tickers and concepts when you request a sample - the extract comes back cut to exactly that shape.