Your bank sends a PDF. Your spreadsheet wants rows. Here is what actually works, what it costs you, and how to tell when the conversion quietly went wrong.
Before anything else, log into your bank and look for an export option — CSV, OFX, QIF or Excel. Most retail banks have one, usually hidden under Statements, Download or Export, and often limited to the last 12 or 24 months.
If it is there, use it and stop reading. A CSV from the bank is the actual data. Everything below is reconstruction — reading numbers back off a printed page — and reconstruction can be wrong in ways an export never is.
You are still here because one of these is true: the statement is older than the export window, it came from a closed account, it arrived by email from a third party, or your bank simply does not offer it. That is the normal case, and it is why this problem exists at all.
| Way | What it costs you |
|---|---|
| Retyping by hand | Time, and transcription errors. A wrong digit in a statement does not announce itself — you find it weeks later when a total does not match. |
| An online converter | You upload the statement to a company you know nothing about: account number, balance, every payee, your address. Free ones fund themselves somehow. |
| A Python script (pdfplumber, camelot) | Nothing, if you write code. Everything, if you do not. |
| A tool that runs in your browser | Nothing leaves the machine, but you have to check the output yourself — which you should be doing anyway. |
Why the upload matters more than people admit. If the statement is yours, uploading it is your risk to take. If it belongs to a client — an accountant, a bookkeeper, a lawyer handling someone's affairs — it is not your risk to take. Sending a third party's financial records to a processor you have not vetted is the kind of thing that sits badly in an audit, and in the EU it is a transfer of personal data you would have to be able to justify.
A PDF does not contain a table. It contains pieces of text, each with a position on the page. Any tool that "extracts a table" is guessing where the columns were, from the gaps. That guess fails in four predictable ways:
12 Mar has a space in it, and a
space wide enough looks like a column break.Whatever tool you use, do this every single time. It takes twenty seconds.
The rule that matters: a converter that fails loudly is worth more than one that fails quietly. If a tool hands you a clean-looking table with one wrong number in it, you have a problem you will not find until it costs you something.
Open the PDF and try to select a line of text with your mouse. If you cannot, there is no text layer: it is an image. No extraction tool can help, because there is nothing to extract. You need OCR first — Adobe Acrobat, Tesseract, or your scanner's own software — and then you are reading numbers a machine guessed from pixels, so the total check above stops being optional.
Tablecut → Drop a PDF, check the table on screen, copy it into your spreadsheet. The file is read inside your browser and never uploaded. Free to use.Written by the people who built Tablecut, a browser tool that does this without uploading your file. © 2026 Tchubi. All rights reserved.