How to convert a scanned bank statement to Excel
A statement downloaded from online banking contains real text: you can select a line and copy it. A scanned statement (a paper statement put through a scanner, a phone photo saved as PDF, or a statement a branch printed and someone scanned back in) contains only a picture of text. To get numbers out of a picture you need either optical character recognition (OCR), which adds a text layer, or a converter that reads the page image directly with a vision model. This guide covers how to tell which you have, both routes, and how to catch the mistakes either can make.
Step 0: Do you actually need to scan?
Before OCR, try to get a digital copy. It is almost always faster and more accurate:
- Most large banks keep years of statements online: Chase up to 7 years depending on account type (Chase), Wells Fargo up to 7 years for deposit accounts (Wells Fargo), NatWest up to 7 years (NatWest), HSBC UK up to 6 years (HSBC UK).
- Older or closed-account statements can often be requested from the bank as PDFs. Bank of America, for example, lets you request copies up to 7 years back, delivered online 24–36 hours later (Bank of America).
If you're a bookkeeper holding a client's scanned pile, ask the client to download the digital versions first. You'll save yourself the OCR clean-up.
Step 1: Check whether your PDF has a text layer
Open the PDF and try either of these:
- Select a transaction line with your cursor. If individual words highlight, there's text. If the whole page highlights as one block (or nothing does), it's an image.
- Search (Ctrl/Cmd + F) for an amount you can see on the page. No result means no text layer.
Some PDFs are hybrids: a scan that someone already ran OCR on. Those have text, but it may contain OCR errors, so treat them like scans when checking.
Step 2: Run OCR
| Tool | Cost | Where it runs | Notes |
|---|---|---|---|
| Adobe Acrobat (Scan & OCR > In this file > Recognize Text) | Paid subscription | On your computer | Can batch multiple files; Adobe recommends reviewing the result and correcting errors (Adobe) |
| OCRmyPDF | Free, open source | On your computer | Adds an OCR text layer to scanned PDFs using the Tesseract engine; can deskew pages first (OCRmyPDF docs) |
| Google Drive (right-click > Open with Google Docs) | Free | Google's servers | Files must be 2 MB or smaller; Google warns that "lists, tables, columns ... are not likely to be detected" (Google Drive Help), so expect a text dump rather than a table |
For bank statements, the tools that keep the original page and add a text layer behind it (Acrobat, OCRmyPDF) are better than ones that produce a new document, because a converter can then use the positions of the words to rebuild columns.
A basic OCRmyPDF run looks like this:
ocrmypdf --deskew scanned-statement.pdf statement-with-text.pdf
Privacy: OCR services that run on someone else's server see every number on the statement. Acrobat desktop and OCRmyPDF run locally.
Step 3: Get the best scan you can
OCR accuracy depends mostly on the image. Adobe's scan settings suggest grayscale or colour at 200–400 dpi (or 400–600 dpi for black-and-white) and offer deskew, background removal and text sharpening to improve recognition (Adobe). Google Drive asks for text at least 10 pixels high, right-side-up pages, and even lighting (Google Drive Help). In practice:
- Flat pages, no folds across the transaction table.
- Straight, not skewed. Run deskew if your tool has it.
- No phone-camera glare or shadows across the numbers.
Step 4: Convert, then distrust every digit
Once there's a text layer, convert as you would a digital statement (PDF to Excel guide). But OCR introduces errors digital PDFs never have:
- Look-alike characters: 0/O, 1/l/I, 5/S, 8/B. A "1,2O0.00" won't add up.
- Lost decimal points and commas: 1,234.56 read as 123456.
- Dropped minus signs or brackets, so a debit turns into a credit.
- Merged columns when the scan was skewed, so the balance digits run into the amount.
This is exactly what a balance check catches. On a scanned statement, run both:
- Foot the statement: opening balance + money in − money out = closing balance.
- Check every row against the printed running balance, if there is one.
A single OCR digit error shows up as a row where the balance stops agreeing. Fix that cell against the paper original. Our reconciliation guide has the Excel formulas.
What about Footed?
Footed reads the text layer of digital PDFs in your browser and sends only that text for parsing. When a page has no text layer, Footed renders it in your browser and sends that page image to a vision model instead. That's slower, and it's the one case where an image of the page leaves your device. If you'd rather keep scans fully local, run OCR first with one of the tools above and convert the OCR'd file. Either way, Footed's reconciliation flags any row where a misread digit broke the running balance.
FAQ
Can I convert a photo of a bank statement? Save it as a PDF and drop it into Footed (it will use the vision route), or run OCR locally first. Phone photos with perspective distortion OCR badly; a flatbed scan or a scanning app that flattens the page is better.
Why are some amounts wrong after OCR but most are right? OCR errors are character-level and random, so one bad digit in a hundred rows is typical. The running-balance check finds it.
Is OCR on a bank statement legal? OCR on your own documents, or ones you're authorised to process (for example as the client's bookkeeper), is just reading them. Don't OCR statements you have no right to.
My scan has handwriting or stamps on it. Will that break things? It can. Annotations near the transaction table may be read as extra characters. Check those rows carefully.
Already have a text-layer PDF? Footed converts it to Excel, CSV or OFX in your browser and highlights every row that doesn't reconcile. Free for 5 pages a day.
Published 2026-10-07.