A scanned bank statement is a picture of text, not text. Before any spreadsheet can use it, the characters on each page must be recognised with OCR (optical character recognition). OCR is impressive but imperfect, so the quality of the scan and a careful review make all the difference.
First, check whether you really need OCR
Try selecting a word in the PDF. If you can, the PDF already contains text and no OCR is needed. If your bank offers statements through online banking, downloading a digital copy is almost always faster and more accurate than scanning a paper one.
How to prepare a good scan
- Resolution: scan at 300 DPI. Lower resolutions blur small digits; much higher adds file size without helping.
- Colour mode: greyscale or black and white is usually clearer than colour.
- Alignment: keep pages straight. Skewed lines make columns harder to detect.
- Lighting: if you use a phone, avoid shadows and glare, and hold the camera parallel to the page.
- One PDF: combine all pages of a statement into a single PDF in the right order.
Converting a scanned statement with FinParse
When you upload a PDF without selectable text, FinParse detects it and offers to run text recognition. The OCR engine (Tesseract) runs inside your browser, so the scanned pages are not uploaded. The first time, the engine files – about 7 MB – are loaded from this website.
Recognition typically takes 10–30 seconds per page depending on your device, and up to 25 pages are processed per file. The recognised text then goes through the same transaction detection used for digital PDFs.
What to check in OCR results
OCR errors follow predictable patterns. Look out for:
- Confused characters: 0 and O, 1 and l or 7, 5 and S, 8 and B.
- Missing or extra decimal points, which change an amount by a factor of 100.
- Minus signs or Cr/Dr markers that were not recognised.
- Rows that were skipped because the date was misread.
The running-balance check is especially useful here: a single misread digit makes a row fail the check, so FinParse marks it for review. Compare flagged rows with the scan, correct them in the table, and confirm that the closing balance matches before downloading.
When OCR is not good enough
Faded, handwritten or heavily stamped statements may not recognise reliably. In that case, request a digital copy from your bank, or use the converter’s table to type the transactions in and export them as Excel or CSV – you still get a clean, consistently formatted file.