Extract data from corporate annual reports

A corporate annual report is the yearly account a company gives its shareholders and its registry, and outside the United States it is not an SEC filing at all. In the United Kingdom a public company lodges its report and audited accounts with Companies House under the Companies Act 2006, a German or Australian company files with its own registry, and each states results under IFRS or a national GAAP rather than the SEC forms. What the reports share is the shape: a business review, the audited financial statements, the auditor's opinion, and the governance section. An equity analyst, a credit team, and a data vendor read the same figures from any of them: total revenue and net income for the year, total assets and shareholders' equity at the year-end, the segment breakdown, the auditor and whether the opinion was clean, and the board and its committees. Its registration number and jurisdiction, not a ticker, identify the entity and set which accounting framework applies. Financial statements are the part that has to reconcile. Consolidated income statements pair the current year against the prior year line by line, the balance sheet does the same for assets and equity, and the segment table has to sum to the group revenue at the top of the report. Alongside the numbers sit the narrative disclosures that a screen still needs as fields: the auditor's name and the audit opinion of unqualified, qualified, adverse, or disclaimer, the significant accounting policies, the principal risks, related-party transactions, and dividends declared. Because a report can run past 200 pages and mixes audited statements with a chairman's letter and a sustainability section, the challenge is pulling the reconciling figures out of the prose and the tables together. Talonic reads the annual report into a structured record that keeps the statements and the disclosures as typed fields. A report from Pennine Industrial Group plc (registration number 04182236, England and Wales), for the year ended 2026-03-31 in GBP, audited by Grant Thornton UK LLP with an unqualified opinion, returns total revenue of 486,300,000 GBP that foots to two segments, net income of 38,700,000 GBP, and the board and its committees, so a coverage model is populated from the report rather than transcribed by hand.

What gets extracted from corporate annual reports

How extraction works for corporate annual reports

Corporate annual reports arrive as designed PDFs, registry filings, and print renders, and because IFRS and national GAAP allow different layouts, the same figure sits in a table on one report and in prose on another. Talonic classifies the report and maps each value to the annual-report field set held in the Field Registry, which separates the company header, the income statement, the balance sheet, the cash-flow statement, the segment table, and the governance blocks rather than flattening the document. Revenue, net income, total assets, shareholders' equity, and the currency are typed, the income statement and balance sheet keep the current and prior year side by side, and the segment table is checked to foot to the group revenue. Auditor name and audit opinion, the accounting policies, the principal risks, and the dividend and related-party disclosures are captured as their own fields, and the board and executive tables keep each member and their role. Per-value confidence and a pixel-region pointer under DIN SPEC 91491 let an analyst verify a revenue figure or an audit opinion against the source. The extraction structures what the report states and does not audit, restate, or advise on the figures.

Sample extraction

A UK PLC annual report and audited accounts for the fiscal year

{
  "document_number": "AR-2026-PIG",
  "document_date": "2026-06-24",
  "fiscal_year_end": "2026-03-31",
  "company.legal_name": "Pennine Industrial Group plc",
  "company.registration_number": "04182236",
  "company.jurisdiction": "England and Wales",
  "company.industry_sector": "Industrial Manufacturing",
  "currency": "GBP",
  "total_revenue": 486300000,
  "net_income": 38700000,
  "total_assets": 612400000,
  "shareholders_equity": 274900000,
  "employee_count": 3120,
  "auditor_name": "Grant Thornton UK LLP",
  "audit_opinion": "unqualified",
  "effective_tax_rate": 21.5,
  "consolidated_statements_of_income": [
    {
      "line_item": "Revenue",
      "current_period_amount": 486300000,
      "prior_period_amount": 452100000
    },
    {
      "line_item": "Cost of sales",
      "current_period_amount": 298100000,
      "prior_period_amount": 281400000
    },
    {
      "line_item": "Operating profit",
      "current_period_amount": 61200000,
      "prior_period_amount": 54800000
    }
  ],
  "segment_revenue": [
    {
      "segment_name": "Industrial Automation",
      "segment_revenue": 312500000,
      "segment_operating_income": 41800000
    },
    {
      "segment_name": "Precision Components",
      "segment_revenue": 173800000,
      "segment_operating_income": 19400000
    }
  ],
  "board_of_directors": [
    {
      "member_name": "Alistair Crewe",
      "position_title": "Chairman",
      "independence_status": "Independent"
    },
    {
      "member_name": "Fiona Mbeki",
      "position_title": "Chief Executive Officer",
      "independence_status": "Management"
    }
  ]
}

Frequently asked