Every transaction, every balance, every page.

Bank statement data extraction turns multi-page PDFs of transaction history into clean, reconcilable structured data. The use case sits at the heart of loan underwriting, cash-flow forecasting, forensic accounting, and any consumer app that aggregates accounts across institutions that don't expose an Open Banking API. Done by hand it's an hour of typing per statement; done with regex it breaks the moment a bank tweaks its layout.

On a typical statement, Ztract extracts the bank name, account holder, account number, and IBAN; the currency and statement period; the opening and closing balance; the summary total debits and total credits; and the full transactions table with date, description, debit, credit, and running balance for every line. The sample above is a 15-transaction HSBC euro account — balances reconcile to the closing total exactly because nothing was dropped or misaligned across the page break.

Statements that run 20+ pages are the norm. Column headers repeat at the top of each page, summary blocks shift between formats, and pending vs cleared transactions live in different sections. Ztract keeps the transaction order intact, deduplicates the repeated headers, and tags pending lines so downstream reconciliation doesn't double-count. Multi-line transaction descriptions — common on wire transfers and SEPA credits — stay attached to the right row.