Converting a scanned bank statement to Excel

There are two kinds of PDF that look identical on screen and behave nothing alike. One was generated by the bank's system and contains real text you can select with your cursor. The other is a photograph of paper wrapped in a PDF container — every character is just pixels.

If you have ever opened a statement, tried to select a line, and watched a blue rectangle drag across the page without highlighting anything, you have the second kind.

Why ordinary tools return an empty file

Most PDF-to-Excel converters extract the text layer. A scanned page has no text layer, so they extract nothing and hand back an empty sheet — often without telling you why. That is the whole explanation for a class of frustrating five-minute detours.

Check which kind you have

Open the PDF and try to select a line of text. If characters highlight, you have a text layer and any decent converter will work. If you can only draw a rectangle over the page, it is scanned and needs a different approach.

What works instead

Scanned pages have to be read as images. StatementSnap renders each page at around 150 DPI and reads it with a vision-capable model, using the same parsing rules as the text path — and, critically, the same arithmetic check afterwards.

That check matters much more here than on a text statement. Reading pixels is inherently less certain than reading characters: a smudged 8 can pass for a 3, and a faint line can be skipped. So the question is not "did the tool produce a table" but "can the table be proven complete."

The proof is the balance equation: beginning balance, plus every transaction, equals the ending balance. If a digit was misread or a line missed, the sum will not close, and you will know immediately rather than three weeks into the quarter.

Practical tips for better scans

• Scan at 300 DPI where you can. It costs nothing and materially improves digit accuracy.
• Scan in greyscale rather than pure black and white — thresholding can erase thin strokes on numbers.
• Keep pages straight. A few degrees of skew is tolerable; a visible tilt is not.
• If the bank still offers the original PDF in online banking, download that instead. A generated PDF beats any scan.

Spot-check the risky rows

Even with a green reconciliation badge, glance at the rows with unusual amounts, and at any date that looks out of sequence. The badge proves the arithmetic closes; it does not prove that a description was transcribed perfectly. Two minutes of scanning the table is worth it on a scanned source.