Bank Statement OCR: From Scanned Pages to Structured Data
FastOCR's bank statement workflow extracts account, balance, and transaction fields into structured JSON and CSV, with XLSX as a best-effort export. It also runs balance consistency and source-grounding checks that produce a review signal. That signal helps route work; it is not a guarantee that every extracted value is correct.
What is bank statement OCR?
Bank statement OCR reads characters from a scanned statement PDF or image. On its own, OCR usually returns page text and may lose the relationship between dates, descriptions, debits, credits, and running balances. Structured bank statement extraction adds a second layer: it maps readable values into account fields and transaction rows, then generates JSON and CSV, with XLSX available on a best-effort basis. FastOCR also checks balance consistency and source-grounding evidence. The result reports whether those checks found an issue, helping a workflow route routine output toward automation or exceptions toward human review. This review signal is not proof that every extracted value is correct. Important amounts should still be compared with the original statement, especially when the scan is blurry, cropped, password-protected, or uses an unusual layout. Start with the bank statement converter, then use the technical reference if you are building a workflow.
What the workflow produces
The extraction is designed for statement-shaped documents, not just a block of copied text. Depending on what is visible in the source, the result can include:
- Statement period, bank name, account details, currency, and opening or closing balances.
- Transaction date, time, description, reference, type, debit, credit, balance, and source page.
- Account totals such as credits, debits, net movement, and transaction count.
- JSON and CSV exports, plus an XLSX export when the workbook conversion succeeds.
Why raw OCR is not enough
A scanned statement is an image of a table. Raw OCR can recognize the text while losing column relationships, decimal points, negative signs, or the connection between a transaction and its running balance. A born-digital PDF may already contain selectable text, but that still does not make it structured financial data. The scanned bank statement OCR page explains the image-quality side; the commercial workflow adds the financial fields and checks.
How the review signal works
The result includes needs_review and a verification object. The checks look for an opening balance anchor, whether statement endpoints tie out, breaks in the running-balance chain, and other flagged issues. A transaction's verified field is based on both arithmetic-chain consistency and source grounding for that row.
| Signal | Meaning |
|---|---|
| needs_review: false | The configured checks found no issue that requires review. It is not a proof of correctness. |
| needs_review: true | Route the flagged account, page, or row to a human and compare it with the original. |
A practical automation-versus-review workflow
- Sign in, upload the statement, and wait for the structured result.
- Use
needs_review,flagged_issues, and row-level signals to separate routine output from exceptions. - For no-issue results, continue only with automation rules that fit your own controls and risk level.
- For flagged results, compare the output with the original statement and correct or approve the affected rows.
- Keep the original statement alongside exported data for any later reconciliation or audit work.
Good fits and clear limits
Useful for
- Preparing bookkeeping and reconciliation work.
- Reviewing spending or cash movement in a spreadsheet.
- Turning statement packs into consistent downstream records.
- Routing exceptions to a person before they enter another system.
Not a substitute for
- Review of a financial record when a decision depends on it.
- Fraud detection, underwriting, or a lending decision.
- A promise that every bank layout or scan will parse the same way.
- Source documents and your own retention, access, and approval controls.
File limits, trial, packs, and retention
Sign-in is required. The trial covers one statement up to 20 pages. PDFs are accepted up to 50 MiB; PNG, JPG, JPEG, and WebP images are accepted up to 20 MiB. Results are retained for about 30 days. Statement packs are one-time purchases valid for 12 months. View current statement tiers and credit packs on our pricing page.
API availability
Programmatic access is in private beta and is allowlisted per account. The API uses an authenticated create, presigned upload, start, poll, and download flow. Read the bank statements API docs for request fields, output schemas, and review signals.
Frequently asked questions
What is the difference between OCR and bank statement extraction?
Raw OCR turns pixels into text. Bank statement extraction also organizes account, balance, and transaction fields into JSON, CSV, or XLSX and adds checks that create a review signal.
Does the review signal guarantee that the statement is correct?
No. It is a workflow signal based on balance consistency and source-grounding checks. Use the original statement for final financial decisions and route flagged results to a person.
Can I try bank statement extraction?
Yes. Sign-in is required, and the trial covers one statement up to 20 pages. PDFs up to 50 MiB and PNG, JPG, JPEG, or WebP images up to 20 MiB are supported.
Is there a bank statement API?
The API is in private beta and is available to allowlisted accounts. The technical reference is at /docs/bank-statements.
Bank statement OCR by industry & vertical
Explore dedicated guides and workflows tailored to specific professional and business use cases:
Accounting & Bookkeeping →
Convert statements to Excel and CSV for QuickBooks and Xero reconciliation.
Mortgage & Loan Underwriting →
Extract borrower deposits, calculate DTI, and audit statement math.
Tax Preparation & CPAs →
Reconcile Schedule C expenses, 1099 income, and year-end client statements.
Legal Discovery & Forensic Accounting →
Trace assets, check numbers, and cash flow for litigation exhibits.
Small Business & Freelancers →
Convert bank statement PDFs to Excel spreadsheets in 3 simple steps.
Real Estate & Property Management →
Verify tenant proof of income and track per-property maintenance ledgers.
Try the bank statement workflow
Sign in to test one statement up to 20 pages, or review the pack options for larger workloads.
Related Articles
Bank Statement OCR for Accounting
Convert client bank statements into Excel and CSV for QuickBooks.
Bank Statement OCR for Mortgage Lenders
Accelerate borrower underwriting and cash flow calculations.
Bank Statement to Excel for Small Business
Convert PDF bank statements to Excel and CSV in 3 simple steps.