Practical guide

How to Convert a Bank Statement PDF to CSV Without Losing the Audit Trail

A useful bank-statement conversion is more than copied text. The resulting spreadsheet must retain every transaction once, preserve exact cents and prove that the extracted rows still agree with the statement.

Start with the original native-text PDF

Download the PDF directly from online banking. A native-text PDF lets a converter read the same date, description and amount text that you can select on screen. A scan or photograph contains pixels instead; OCR can introduce digits that look plausible but are wrong.

Keep the original statement unchanged with your normal workpapers. Conversion should create a working file, not replace the source evidence.

Use a CAD chequing statement, not a credit card or loan statement.
Do not print and rescan a downloaded PDF.
Reject password-protected, damaged or incomplete files rather than bypassing controls.

Check completeness before cleaning descriptions

The first test is whether every statement page and transaction table can be accounted for. Printed page numbering, opening and closing balances, transaction counts, deposit totals and withdrawal totals provide independent evidence.

A spreadsheet that balances can still be incomplete when two omitted rows offset one another. That is why a converter should use every available printed control and retain page references instead of relying on one equation.

Opening balance + money in − money out must equal closing balance.
Printed totals and counts should agree where present.
A complete running-balance chain should agree row by row.

Preserve rows, order and exact cents

Amounts should be parsed as exact decimal cents, not floating-point approximations. Same-day rows must stay in statement order, and two rows that look identical must remain two rows unless a reviewer explicitly documents an exclusion.

Descriptions may wrap across lines. Join legitimate continuation text without translating customer text, removing reference digits or turning HTML-looking text into markup.

Keep source references and reviewed corrections

A page number and source location shorten review: select a spreadsheet row and compare it with the highlighted statement line. If a correction is necessary, keep the extracted original, the replacement, the reason and an exact undo path.

Never plug a balance, infer a missing transaction or change the sign merely because the result then reconciles. An ambiguous file should remain blocked.

Export for the next job, not just the current screen

A canonical CSV should use explicit headers, ISO dates, separate money-out and money-in columns, literal descriptions and CRLF line endings. Excel should contain real date and numeric money cells. Reopen the outputs and compare their row counts and totals before using them.

One Purpose Tools performs these checks locally for supported statement structures. The free review shows whether a document qualifies before you buy the reusable package.