Financial statement OCR: balance sheets and P&Ls straight into Excel.
Audited packages, reviewed statements, scans and phone photos, from every preparer. ParserBee reads each line with both period columns and gives you rows you can tie out and spread in your own sheet.
ParserBee, Excel's PDF import or spreading software?
Excel's import is free and fine for one clean PDF. Spreading platforms do the credit work too, at a bank's price. ParserBee sits between them.
| Excel Get Data From PDF | Spreading software | ||
|---|---|---|---|
| Scans and phone photos | ParserBeeYes: PDF, JPG, PNG, WebP and HEIC | Excel Get Data From PDFReads tables in PDFs; Microsoft's documentation does not cover scanned images | Spreading softwareUsually, with a review step |
| Many preparers' layouts | ParserBeeOne Parser Template reads them all | Excel Get Data From PDFEach file's tables need cleaning in Power Query | Spreading softwareBuilt for it |
| Line items with two periods | ParserBeeOne row per line, both columns | Excel Get Data From PDFWhatever table it detects | Spreading softwareYes |
| Standard spread and ratios | ParserBeeIn your own sheet, set up once | Excel Get Data From PDFIn your own sheet | Spreading softwareBuilt in, with covenants |
| Cost | ParserBee25 pages free, then from $29 a month for 250 pages | Excel Get Data From PDFIncluded with Excel | Spreading softwareQuoted per contract |
| Best for | ParserBeeAccountants, analysts, lenders and finance teams who spread in Excel | Excel Get Data From PDFOne clean, digital PDF | Spreading softwareBanks spreading thousands of borrowers |
Excel column: from Microsoft's documentation for Get Data From PDF. Spreading software column: from the spreading vendors' own sites, where prices are quoted per contract. Both checked November 2026; check their sites for current details.
Spreading starts with retyping, and retyping is where the errors come from.
Before anyone can look at a ratio, someone types two years of statements into a spreadsheet. That is the part OCR in finance should take away.
Every preparer, a new layout
One CPA firm's balance sheet looks nothing like the next. Line names, order, subtotals and column headings all change.
Hours of retyping per file
Two years of balance sheet and income statement lines, typed one cell at a time, then checked again because one slip breaks the totals.
Scans, photos and long PDFs
Statements arrive as scanned packages, phone photos and 40-page annual reports. Copy and paste from a PDF loses the columns.
Every line read, then tied out to zero.
Upload one statement and the AI template builder proposes the fields in about 20 seconds: company, statement type, period, unit scale, and one row per line item with both period columns. Your sheet then checks that the numbers add up, as on this fictional balance sheet.
| 2025 | 2024 | |
|---|---|---|
| Assets | ||
| Cash and cash equivalents | 1,240 | 980 |
| Accounts receivable, net | 2,315 | 2,040 |
| Inventories | 1,870 | 1,655 |
| Prepaid expenses | 145 | 120 |
| Total current assets | 5,570 | 4,795 |
| Property and equipment, net | 3,410 | 3,560 |
| Total assets | 8,980 | 8,355 |
| Liabilities and equity | ||
| Current portion of long-term debt | 300 | 300 |
| Total current liabilities | 2,040 | 1,815 |
| Long-term debt | 2,100 | 2,400 |
| Total liabilities | 4,140 | 4,215 |
| Total equity | 4,840 | 4,140 |
| Total liabilities and equity | 8,980 | 8,355 |
A misread figure would leave the difference above zero and point you to the section to check. Two formulas do this for every statement you read.
How do I get financial statement numbers from PDFs into a spreadsheet without retyping?
Run the statements through a document parser that returns one row per line item with both period columns, then let formulas tie out the totals and map the lines to your spread. For public US companies, skip the PDF: the SEC publishes the numbers as data.
Check whether the numbers already exist as data
US public companies file in XBRL, and the SEC publishes the face financials as free quarterly Financial Statement Data Sets. For one clean digital PDF, Excel's Data, Get Data, From File, From PDF can pull a table. Private company statements, scans and many layouts are where OCR earns its keep.
Build the template from one statement
Upload one balance sheet or income statement to the AI template builder. Keep fields like
company_name,statement_type,period_end,currencyandunit_scale, and a list ofline_itemswithsection,line_item,current_periodandprior_period. Add plain-English rules to the descriptions, such as "amounts in parentheses are negative".Read the statements
Drop up to 100 files into Data Lab and download one Excel file, one row per line item; call that sheet Rows. PDFs are read up to 24 pages per file on paid plans, so upload the statement pages of a long annual report, not the notes. Fill the company and statement columns down, as the export puts them on each statement's first row.
Tie out the totals
A misread number shows up as a difference. With line labels in column D and the current period in column E:
=SUMIFS(Rows!E:E, Rows!D:D, "Total assets") - SUMIFS(Rows!E:E, Rows!D:D, "Total liabilities and equity")
Do the same for each section: the lines of current assets should add up to total current assets.
Map the lines to your spread
Keep a Map sheet that turns each preparer's label into your standard line ("Notes payable to bank" and "Line of credit" both become "Short-term debt"). Add the mapped line to Rows in column G with XLOOKUP, then sum it into the spread and calculate the ratios:
=XLOOKUP(D2, Map!$A$2:$A$500, Map!$B$2:$B$500, "UNMAPPED") =SUMIFS(Rows!$E:$E, Rows!$G:$G, A5)
Any line marked UNMAPPED is new to your map: add it once and the next statement from that preparer maps itself.
From statement rows to a checked spread.
ParserBee is not a credit platform. It reads the statements; the checks, the mapping and the ratios live in your sheet, where you can see every rule. This is Kestrel Components' fiscal 2025, from the balance sheet above and its income statement: revenue 14,620, net income 850 (in thousands).
- Balance check: 8,980 of assets against 8,980 of liabilities and equity
- Current ratio: 5,570 / 2,040 = 2.73
- Debt to equity: (300 current + 2,100 long-term) / 4,840 = 0.50
- Net margin: 850 / 14,620 = 5.8%
=SUMIFS(Rows!E:E, Rows!D:D, "Total assets") - SUMIFS(Rows!E:E, Rows!D:D, "Total liabilities and equity")| A | B | C | |
|---|---|---|---|
| 1 | Spread line | FY2025 ($000) | Check |
| 2 | Revenue | 14,620 | |
| 3 | Net income | 850 | |
| 4 | Total current assets | 5,570 | |
| 5 | Total current liabilities | 2,040 | |
| 6 | Total debt | 2,400 | |
| 7 | Total equity | 4,840 | |
| 8 | Assets = liabilities + equity | 0 | OK |
| 9 | Current ratio | 2.73 | |
| 10 | Debt to equity | 0.50 | |
| 11 | Net margin | 5.8% |
When you need a spreading platform, or no OCR at all.
- The company is listed in the US: use the SEC's XBRL data, which is already structured and free.
- You need a standard chart, covenant tests and credit memos in one system: choose a spreading platform.
- You want the tool to judge the numbers: ParserBee reads them; it does not analyse or approve.
- One file is a 200-page annual report: upload the statement pages; ParserBee reads up to 24 pages per PDF on paid plans.
Who spreads faster with it.
- Accounting firms taking in client statements from other preparers.
- Community lenders and brokers who spread borrowers in Excel.
- Analysts comparing private companies from scanned packages.
- Finance teams consolidating subsidiaries' statements.
From the shared inbox to your spreading workbook.
Keep your spreading model and your loan or accounting system. ParserBee sits between the document inbox and your workbook: the rows go into Excel, Google Sheets or Airtable, or anywhere else through Zapier, Make, n8n, a CSV file or the API.
Building it yourself? The document parsing API returns the same fields as JSON, with Python and Node.js libraries.
Bank statements, invoices and tax forms go the same way.
83 ready-made document parser templates across 16 industries, plus any document you build a template for.
Bank statements
Every transaction from a PDF statement into Excel or CSV, free and without signup.
PDF bank statement to Excel →Credit card statements
Card transactions, dates, merchants and amounts.
Credit card statement extraction →Financial statement OCR questions
What is financial statement OCR?
It is software that reads a balance sheet, income statement or cash flow statement from a PDF, scan or photo and returns the numbers as data. Plain OCR gives you text; financial statement OCR keeps each line with its label and its period columns, so it lands in a spreadsheet ready to check and analyse.
Can I automate financial spreading from PDF statements?
Partly, with ParserBee, and fully with dedicated spreading software. ParserBee reads every statement line, with both period columns, into Excel or Google Sheets. A mapping table in your sheet then sums the lines into your standard spread and calculates the ratios, as shown on this page. Banks that spread thousands of borrowers and track covenants use spreading platforms that do the mapping and the credit analysis too.
Does it work as OCR for balance sheets and income statements?
Yes. One Parser Template reads balance sheets, income statements and cash flow statements, with a field that says which statement each line came from. It reads each line label, the current and prior period amounts, and the subtotals, from digital PDFs, scans and phone photos.
How do I convert audited financial statements to Excel?
Upload the statement pages to ParserBee and download Excel, one row per line item with both years. An audited package often runs to 20 or 30 pages with the notes; ParserBee reads PDFs up to 24 pages per file on paid plans (12 on the free trial), so upload the pages with the primary statements, or split the file. For public US companies, the SEC already publishes the numbers as data, so start there.
Is there a ready-made financial statement template?
Not yet. Build one in about 20 seconds with the AI template builder: upload one statement (PDF, PNG, JPG or WebP, up to 15 MB) and it proposes the fields, including the list of line items. Check them, run the test, and the same Parser Template reads every other preparer's statements.
How does it handle negative numbers, scale and multi-year columns?
Describe the rules in plain English in the field descriptions, for example that amounts in parentheses are negative and that the prior period column is the right-hand one. Add a field for the unit line, such as "in thousands", and multiply in your sheet. Each line comes back with its current and prior period amounts in separate columns.
How do I know the numbers are right?
Check the totals, not every cell. In your sheet, total assets minus total liabilities and equity should be zero, and the lines of each section should add up to its subtotal. A SUMIFS on the extracted rows does both, so a misread number shows up as a difference you can trace to one line.
Can I get public company financial statements without OCR?
Yes. US public companies file their statements with the SEC in XBRL, and the SEC publishes the numbers from the face financials as free quarterly Financial Statement Data Sets. OCR is for the statements that are not in EDGAR: private companies, borrowers, clients and scanned packages.
Does it do ratio analysis or credit decisions?
No. ParserBee reads the statements. Ratios, covenant tests and credit decisions happen in your spreadsheet or your lending system, with your own rules.
Can it read bank statements too?
Yes. The free bank statement converter turns a PDF bank statement into Excel rows with no signup, and the same account reads statements in bulk.
How much does it cost?
New accounts get 25 free pages, no credit card. Paid plans start at $29 a month for 250 pages; see pricing. Each page uses one page credit, so a two-page balance sheet and income statement uses two.
Is my data used to train AI?
No. Your documents are never used to train AI models, and files travel to ParserBee over an encrypted (HTTPS) connection.