bank-statement-reconciliation

Compare three things: row counts, the closing balance extracted from the PDF header against the last CSV Balance value, and chain integrity where this month's opening equals last month's closing. If closing balance and chain pass, the data is usable even with minor row-count differences.

Run pdftotext with the -layout and UTF-8 flags, then apply a regex like CLOSING BALANCE.*?([\d]+\.\d{2}) to capture the amount. Compare it to the final Balance column value in the CSV to the cent.

Many banks' annual or lumped exports silently drop rows that exist in monthly files. Wio Business annual CSVs are documented to drop transactions including inflows. Always use monthly YYYY-MM.csv files for line-item work and treat annual files only as closing-balance checks.

PDF text extraction can over-count because wrapped multi-line transaction descriptions produce extra lines starting with date-shaped strings. Off-by-one-or-two differences are usually extraction artifacts, not real gaps. Trust the closing balance over the line count.

Column names vary by bank: Wio uses Balance, while others use Running Balance or Closing Balance. Avoid hardcoding names and instead locate columns with header.index() plus a fallback so the script works across banks.