Year-end catch-up bookkeeping from bank statements: a practical method

A repeatable process for rebuilding a year of books from PDF bank and card statements: collect, convert, verify, categorize, reconcile.

Catch-up bookkeeping, meaning rebuilding months or years of records from statements, is mostly an extraction problem followed by a categorization problem. Here is a method that scales from one account to a dozen.

1. Collect

Download every statement as a PDF from online banking. Most banks keep 18–84 months online; anything older may require a request to the bank (often with a fee). Name files consistently: bank-account-YYYY-MM.pdf.

2. Convert in batch

Drop all statements for one account into Rowledger. With Pro, they process in batch, each with its own reconciliation badge. Statements that do not reconcile get attention first. Usually that is a scan with one misread row, fixed in seconds in the table.

3. Verify continuity

Closing balance of month N should equal opening balance of month N+1. A gap means a missing statement. The merged export sorted by date makes this easy to eyeball: the Balance column should flow continuously.

4. Categorize

Export with categories enabled (Pro). Rowledger applies keyword rules (merchant → category) as a first pass; review the uncategorized rows and add your own rules for recurring payees. Then import into your accounting software and finish categorization there, or keep the spreadsheet as the ledger for very small entities.

5. Reconcile in the accounting software

Because each converted statement reconciled to its printed balances, the bank reconciliation in QuickBooks/Xero should close with zero difference month by month. If it doesn't, the discrepancy is in categorization or duplicate imports rather than in the extraction, which is a useful narrowing.

Time expectations

A year of monthly text-PDF statements for one account converts in about a minute. Scanned statements add a few seconds per page. The categorization review is where the human time goes; the extraction shouldn't be.