Extract data from SEC Form 10-Q filings

Four times a year a public company reports to the SEC, and three of those reports are 10-Qs, the quarterly filings that keep the market current between annual reports. A 10-Q is lighter than a 10-K, its financial statements are unaudited, and it covers the three-month and often the nine-month periods, but the same readers depend on it. A buy-side analyst updating a model, a lender testing a covenant, and a data vendor feeding a screen all read the quarter's revenue, margins, cash, and share count from the same statements. The quarterly cadence is exactly what makes a 10-Q easy to misread. Figures are stated for both the three months and the nine months ended on the same date, and mixing the two overstates a quarter's revenue nearly threefold. The statements still have to balance, so total assets equal total liabilities plus equity at the period end, and gross profit equals revenue less cost of revenue for the period. Segment revenue is restated for the quarter, EPS is reported basic and diluted for the period, and the comparative prior-year quarter sits in an adjacent column that a parser must not blend into the current one. Talonic reads the 10-Q and returns the registrant, the period, and the quarter's income statement, balance sheet, and cash flow statement as typed fields, with the three-month and nine-month columns kept apart. For Northwind Robotics Inc. (CIK 0001234567), the quarter ended 2026-09-30 in USD reports revenue of $812,000,000 less cost of revenue of $447,000,000 for a gross profit of $365,000,000, net income of $83,000,000 for a basic EPS of $0.55, and a period-end balance sheet where total assets of $3,700,000,000 tie to total liabilities of $2,100,000,000 plus equity of $1,600,000,000. The quarter loads into a model with the periods labeled.

What gets extracted from SEC Form 10-Q filings

How extraction works for SEC Form 10-Q filings

A 10-Q comes off EDGAR as an HTML or PDF filing that states most figures twice, once for the three months ended and once for the nine months ended on the quarter-close date. Talonic classifies the filing and maps it to the financial-statement schema in the Field Registry, which models the US-GAAP concepts, and it tags each amount with the period it belongs to so a three-month revenue is never blended with a nine-month figure. The period-end balance sheet is checked for the assets-equals-liabilities-plus-equity identity, gross profit is reconciled to revenue less cost of revenue for the period, and the comparative prior-year quarter is stored as its own period. Business-segment revenue and operating income are kept per segment, and the management discussion is captured as text. Provenance for each field follows DIN SPEC 91491, pairing a confidence score with a pixel region, so a covenant test or a quarterly model can be checked against the source filing before it is trusted.

Sample extraction

A third-quarter SEC Form 10-Q, three-month figures

{
  "document_number": "0001234567-26-000041",
  "document_date": "2026-10-29",
  "period_end_date": "2026-09-30",
  "fiscal_period": "Q3",
  "fiscal_year": 2026,
  "entity_name": "Northwind Robotics Inc.",
  "company": {
    "cik": "0001234567",
    "ticker": "NWRB"
  },
  "reporting_currency": "USD",
  "total_revenue": 812000000,
  "cost_of_revenue": 447000000,
  "gross_profit": 365000000,
  "research_and_development_expense": 104000000,
  "selling_general_administrative_expense": 142000000,
  "operating_expenses": 246000000,
  "operating_income": 119000000,
  "interest_expense": 14000000,
  "pretax_income": 105000000,
  "income_tax_expense": 22000000,
  "net_income": 83000000,
  "weighted_average_shares_basic": 151000000,
  "earnings_per_share_basic": 0.55,
  "earnings_per_share_diluted": 0.54,
  "cash_and_cash_equivalents": 910000000,
  "total_assets": 3700000000,
  "total_liabilities": 2100000000,
  "stockholders_equity": 1600000000,
  "total_liabilities_and_equity": 3700000000,
  "cash_flow_from_operations": 168000000
}

Frequently asked