Diagnose a bank PDF before choosing a table extraction tool

Before comparing converters, check what your PDF actually contains. A file can look crisp while being only an image, or contain searchable text whose reading order does not match the visible columns. “PDF” describes the container, not the reliability of its transaction table.

This is a preparation checklist, not an upload tutorial for this website. BankStatementScan currently offers a synthetic review example, with no document processing. Use your authorised local viewer to examine your own file; do not share a financial document merely to run this diagnosis.

Test more than the first page

Try selecting a short transaction description and copying it into an empty local text document. Inspect the characters and their order. If selection is impossible, the page may be image-based or restricted. If selection works, that does not establish that the PDF was born digital: a scan can have a text layer added by OCR.

Repeat the check on a middle page and the last page. A document may combine original pages and scanned inserts. Record the pages requiring a different method instead of treating the entire file as one uniform input. Check missing edges, orientation, page numbers and whether any transaction continues across a page break.

Match the tool to the evidence

Tabula is a free, open-source option for table selection and preview in text-based PDFs; its documentation excludes scanned documents. This makes it a useful alternative when the text is present and the table can be selected correctly. Inspect the preview rather than assuming selection guarantees clean rows.

Microsoft documents a PDF connector in Power Query. Availability depends on the host/version. Its guidance includes large-file and multiline-row considerations. These are options to investigate with your installed software, not guarantees about a particular bank layout.

An image-only page needs a text-recognition step somewhere in the workflow. This guide does not choose an OCR provider or promise what a scan can recover. A blurred decimal point or cropped amount may require a better source file rather than another conversion attempt.

Inspect a boundary, not just a pretty row

Choose three difficult locations: a description split over two lines, the last transaction on a page, and the first row after a repeated header. Check that the description stays with the right date and amount. Ensure the repeated header, subtotal or balance carried forward does not become another transaction.

Make a small review record: source page, text present or absent, problem observed, and proposed method. Preserve the original. Never change an amount in the source to help an extractor. If you make a separate working copy, keep its relationship to the original clear.

Decide whether conversion is necessary at all

Check whether the bank already provides the desired period as CSV or another structured format. If it does, verify its columns and dates before choosing the PDF route. There is no reason to introduce a recognition step just because a specialised converter exists.

When you do extract, move from document diagnosis to balance review and the planned Excel/CSV workflow. These checks address different errors; none alone certifies the result.

You can request access through this website’s form. It sends your email and visit context to BankStatementLab for BankStatementScan; a recorded request does not create an account or provide immediate access. See privacy.