How to redact or anonymize bank and card statements (PDF)
Bank and card statements are needed for dispute and fraud workflows, lending and underwriting tests, parser QA and vendor evaluations. They carry the account holder's name and address, account and card numbers, routing numbers and transaction details.
Redaction or anonymization?
Redaction draws black boxes. It is right when a document is being released and the values must simply disappear. It is wrong for testing: a parser cannot extract an account number that is no longer there, and an agent cannot match a statement to a dispute form.
Anonymization by synthetic replacement swaps each identifying value for a realistic fake one, in place. The statement still looks and parses like a statement. This is what you want for test data, demos and evals.
What to replace on a statement
- Account holder names and mailing addresses
- Account numbers, card numbers (PANs) and last-four references, IBAN and routing numbers
- Card verification codes, PINs or track data if they appear (these should be replaced with fresh random digits, never derived from the original)
- Phone numbers and email addresses
- Merchant names often matter for disputes and categorization; check in review whether they are kept or replaced for your use
Card numbers should be replaced with checksum-valid numbers, and the same account must map to the same replacement on every page and in every related document (dispute letter, receipt, ID).
Doing it with CasePatch
Upload the statement PDF at the CasePatch home page, or open the public CFPB sample statement there to see every change. CasePatch rewrites values in the PDF text in the original font and position, so the result is still a text PDF. Review the change list, override anything, and export. Free: 10 pages a day without an account, 100 pages a month with one.
Use test or specimen statements in the hosted beta. CasePatch does not certify PCI DSS compliance, and automated detection can miss values: review every page. Related: redaction vs. anonymization, eval sets from production documents.