Bank statement OCR
Optical character recognition built for statement tables, not prose. Upload a scan or a phone photo and get dates, descriptions, debits, credits and a reconciled running balance — ready for Excel, CSV or your accounting software.
Table-aware, not text-dump
Generic OCR returns a wall of characters. Column detection rebuilds the actual grid so amounts stay with their dates.
Balance-checked output
Every recognised row is validated against the running balance, so a mis-read digit is flagged instead of silently exported.
Scan, photo or fax
De-skew and contrast normalisation handle angled phone shots, grey photocopies and low-DPI faxes.
How to OCR a bank statement
Three steps, about a minute for a typical 12-page statement.
- Step 1
Upload the scan or photo
Accepts scanned PDFs, phone photos, and screenshots of a statement page. Skewed and slightly rotated pages are de-skewed automatically.
- Step 2
OCR and table reconstruction
Each page is recognised character by character, then the transaction grid is rebuilt from the column positions — dates, descriptions, debits, credits and the running balance.
- Step 3
Verify and export
The running balance is recomputed and mismatched rows are highlighted for review before you export to Excel, CSV or an accounting format.
When you need OCR — and when you don't
OCR needed
- · Statement scanned at a branch or on a home scanner
- · Photo taken with a phone camera
- · Passbook pages or printed counterfoils
- · PDFs produced by faxing or photocopying
OCR not needed
- · PDFs downloaded from internet banking
- · Statements emailed by the bank
- · CSV, OFX, QFX or QIF exports
- · Anything where you can select the text in a PDF reader
Frequently asked questions
What does OCR stand for?
Optical character recognition. It is the technology that turns pictures of text — a scan, a photo, a fax — into characters a computer can search, copy and calculate with.
How accurate is OCR on bank statements?
Generic OCR gets individual characters right but loses the table structure. A statement-aware pipeline adds column detection and a running-balance check, which is what lifts usable accuracy from roughly 80% to the high nineties on clean scans.
Does OCR work on phone photos?
Yes, provided the whole page is in frame and in focus. Shoot straight on in even light; the de-skew step handles moderate angles.
Is OCR needed for a downloaded PDF?
Usually not. A PDF downloaded from internet banking has a text layer that can be read directly, which is faster and more accurate than OCR. OCR kicks in automatically only when no text layer is present.
Can OCR read handwritten entries?
Handwritten amounts in a passbook-style statement are recognised at lower confidence and flagged for manual review rather than exported silently.
What is a bank statement?
A bank statement is a periodic document issued by your bank listing every transaction on an account for a period — usually one month. It shows the opening balance, each debit and credit with a date and description, and the closing balance. Banks issue it as a PDF, a printed copy, or a downloadable CSV/OFX file.
How do I get a bank statement?
Sign in to your bank's internet banking or mobile app, open the account, choose Statements (sometimes 'e-statements' or 'Documents'), pick a date range, and download the PDF. Most banks keep 12–24 months online; older periods are requested from a branch or support. You can upload the downloaded PDF here directly — no need for the CSV version.
How long should I keep bank statements?
As a rule of thumb keep them seven years if they support a tax return, three years for general business records, and at least one year for personal budgeting. Digital copies count in most jurisdictions, so converting statements to Excel or CSV is a valid way to archive them in a searchable form.