How to convert a scanned bank statement to Excel

Here's a situation that comes up a lot. You run a statement through a PDF to Excel converter, and you get back an empty spreadsheet. Or one cell with the bank's logo text in it. Or nothing at all.

The converter isn't broken. Your statement is a picture.

Text PDF or scanned PDF? Ten-second test

Open the statement and try to select a single transaction amount with your cursor.

• The text highlights: it's a text PDF. The words are stored as real characters. Almost any converter can read it.
• Nothing highlights, or the whole page highlights as one block: it's a scan, or a PDF made from photos. The page is an image of text, not text.

Scanned statements are more common than you'd expect. Older statements requested from a branch, statements a client photographed on their phone, anything that went through a fax or a scanner on its way to you. Some banks even generate PDFs that are images to make them harder to edit.

Why scans are harder

To get transactions out of a scan, something has to look at the image and recognize the characters. That's OCR (optical character recognition). The trouble is that bank statements are close to a worst case for it:

• Small print, dense rows, and thin table lines.
• Numbers that look alike when the scan is fuzzy: 3 and 8, 1 and 7, 5 and 6.
• A decimal point that gets lost, turning 12.50 into 1250.
• Rows that drift slightly as they go down the page, so the amount gets attached to the wrong description.

That's not a reason to avoid scans. It's a reason to check the result properly, which I'll come back to.

Ways to do it

Excel's "From Picture". Newer versions of Excel can read a table from an image: Data > From Picture > Picture From File (or From Clipboard if you take a screenshot). It works one image at a time, so you screenshot each page, import it, review the cells it wasn't sure about, and insert. On Windows it needs Windows 10 version 1903 or later. Fine for a page or two. Tedious for a twelve-page statement.

Google Drive. Upload the PDF, right-click, and open it with Google Docs. Docs runs OCR and gives you the text. You'll get the characters, but usually not the table, so you'll still be rebuilding rows and columns by hand.

Adobe Acrobat Pro. It can run OCR on a scanned PDF to make it searchable, then export to Excel. The OCR is decent. The table detection on statements is hit and miss.

A bank statement converter that reads images. BankStatementHub reads every page as an image anyway, so a scanned statement goes through the same process as a downloaded one. You pick the pages, it extracts date, description, amount and balance for each row, and you review it side by side with the original page before downloading.

Get a better scan, get better results

If you're the one doing the scanning, a few minutes here saves a lot of fixing later:

• Scan at 300 DPI. Lower and small digits blur together. Much higher just makes huge files.
• Keep the page straight. A slightly tilted scan can shift rows into each other.
• Use black and white or grayscale, not colour. It's cleaner for text.
• Flatten the page. Creases and curled edges at the binding throw OCR off.
• Phone photos: use a scanning app (most phones have one built into the notes or files app) rather than the regular camera. It flattens and straightens the page for you.

How to check the result

With a scanned statement, assume there's at least one mistake until you've proven there isn't. The fastest way to prove it is the running balance.

Most bank account statements print a balance after every transaction. Each row should satisfy one simple rule:

previous balance + this transaction = this balance

If a 3 was read as an 8, or a decimal point disappeared, that row breaks the rule, and every other row still passes. So instead of rechecking 200 numbers, you only look at the one or two that fail.

BankStatementHub runs this check automatically and flags the rows that don't add up. If you're working in Excel yourself, add a column with the formula and filter for anything that isn't zero.

Credit card statements usually don't print a running balance, which makes them harder to verify. For those, at least check that the total of all transactions matches the statement's totals for purchases and payments.

The short version

1. Test whether the PDF has real text by trying to select an amount.
2. If it's a scan, use something that does OCR and understands tables, not a generic PDF converter.
3. Scan cleanly if you can: 300 DPI, straight, grayscale.
4. Check the result with the running balance, not by eyeballing every row.