Extract data from SEC Form 10-K filings
A 10-K is the annual report a public company files with the SEC, and it is the single richest disclosure a US issuer produces. Analysts, credit teams, and data vendors pull the same structure from every one: the registrant's name and Central Index Key, the fiscal year covered, the audited balance sheet, the statement of operations, the statement of cash flows, and the narrative sections for risk factors and management's discussion. The financial statements are tagged in XBRL against the US-GAAP taxonomy, which is what lets a spreadsheet compare one filer to another, but the filing itself still arrives as a long PDF or an HTML document on EDGAR. Pulling clean numbers out of a 10-K is harder than it looks because the statements interlock and have to balance. Total assets must equal total liabilities plus stockholders' equity; gross profit is revenue less cost of revenue; net income flows from pretax income less the tax provision and then into retained earnings and the cash flow statement. A single figure such as net income appears in several places and has to reconcile everywhere it lands. Prior-year columns sit beside the current year, segment tables restate revenue by business line, and footnotes carry the detail that changes how a headline number reads. Talonic reads the 10-K and returns the registrant identifiers, the balance sheet, the income statement, and the cash flow statement as typed numeric fields, with the accounting identities checked. For Northwind Robotics Inc. (CIK 0001234567, ticker NWRB), fiscal year ending 2025-12-31 in USD, total assets of $3,500,000,000 tie to total liabilities of $1,900,000,000 plus stockholders' equity of $1,600,000,000; revenue of $2,980,000,000 less cost of revenue of $1,640,000,000 gives a gross profit of $1,340,000,000; and net income of $304,000,000 over 152,000,000 basic shares gives a basic EPS of $2.00. The statements load into a model instead of a retyped page.
What gets extracted from SEC Form 10-K filings
How extraction works for SEC Form 10-K filings
A 10-K arrives from EDGAR as an HTML filing or a PDF, and while the US-GAAP concepts are standard, every issuer names and orders its line items a little differently. Talonic classifies the filing and maps it to the financial-statement schema in the Field Registry, which mirrors the US-GAAP statement concepts, so total revenue, operating income, and total assets bind to their canonical fields regardless of the label the company used. The balance-sheet identity is checked so total assets equal total liabilities plus equity, gross profit is reconciled to revenue less cost of revenue, and net income is traced from the income statement into retained earnings and the top of the cash flow statement. Prior-year columns and segment tables are kept as separate periods rather than merged. Each value returns with a confidence score and a pixel-region pointer aligned with DIN SPEC 91491, so an analyst can verify a reported figure against the source filing before it enters a model.
Sample extraction
A full-year SEC Form 10-K for a mid-cap issuer
{
"document_number": "0001234567-26-000012",
"document_date": "2026-02-18",
"fiscal_year_end": "2025-12-31",
"entity_name": "Northwind Robotics Inc.",
"company_cik": "0001234567",
"company_ticker": "NWRB",
"reporting_currency": "USD",
"fiscal_year": 2025,
"total_revenue": 2980000000,
"cost_of_revenue": 1640000000,
"gross_profit": 1340000000,
"research_and_development_expense": 380000000,
"selling_general_administrative_expense": 520000000,
"operating_expenses": 900000000,
"operating_income": 440000000,
"interest_expense": 55000000,
"pretax_income": 385000000,
"income_tax_expense": 81000000,
"net_income": 304000000,
"weighted_average_shares_basic": 152000000,
"earnings_per_share_basic": 2,
"earnings_per_share_diluted": 1.96,
"total_assets": 3500000000,
"total_liabilities": 1900000000,
"stockholders_equity": 1600000000,
"total_liabilities_and_equity": 3500000000,
"retained_earnings": 1050000000,
"cash_flow_from_operations": 520000000,
"capital_expenditures": -260000000,
"auditor_name": "Ashford & Vance LLP",
"auditor_opinion": "unqualified",
"internal_controls_effective": true
}Frequently asked
Does it check that the balance sheet balances?
Yes. Total assets are reconciled against total liabilities plus stockholders equity, and a filing where the two sides do not tie is flagged rather than returned as clean data.
How does it handle prior-year and segment columns?
Each reporting period is kept separate, so the current fiscal year and the comparative prior year are distinct records, and segment tables that restate revenue by business line are captured as their own rows.
Are the income statement and cash flow statement linked?
Net income is traced from the statement of operations into retained earnings on the balance sheet and into the top of the cash flow statement, so the figure reconciles across the three statements.
Does it capture the auditor opinion?
The independent auditor name and the opinion type (unqualified, qualified, adverse, or disclaimer) are captured, along with management conclusion on internal controls over financial reporting.
Ready to extract from your own SEC Form 10-K filings?
Author note
Reviewed by Talonic engineering · last reviewed 2026-07-06