Skip to main content
FinanceConvert.
CONVERSION QUALITY

A file that opens can still contain errors.

Check the financial meaning of the result, not just its extension. A successful conversion means the software produced an export; it is not independent proof of completeness or importer compatibility.

Reviewed 19 October 2026 · Scope and verification method

Structured files and documents need different readers

OFX, QFX, QBO, QIF, IIF and CSV contain transaction fields the service reads directly. PDFs contain page layout. The local PDF reader supports simple dated lines with signed amounts; scans, separate debit/credit columns and complex tables may require optional Mistral document extraction. Cloud extraction is disabled unless configured and explicitly allowed by an eligible user.

Compare the source page by page

  1. Count transactions, including records near page boundaries and multiline descriptions.
  2. Confirm the first and last transaction dates and the source date convention.
  3. Compare debit and credit totals separately for each currency and account.
  4. Check a few amounts with decimal/thousands separators and any reversals or refunds.
  5. Review running-balance warnings. A clean balance check does not prove completeness.

Cases that deliberately stop the conversion

Conflicting CSV amount columns, repeated column headers, unfinished QIF transactions, ambiguous dates and unsupported investment lists produce errors instead of silently dropping or repairing rows. Bank outputs reject mixed accounts/currencies and repeated transaction IDs. Re-export per account or choose a supported tabular workflow.

Bank imports need identity and mapping

QBO/QFX eligibility depends on the destination product and institution identifiers. IIF and Tally XML use account/ledger classifications that require review. Try imports with a copy or test company. Missing identity and default classifications are not a valid accounting decision.

How retained examples can improve the reader

The service retains private encrypted source/result copies for 90 days, including failures. Comparing future extraction versions with human-checked expected transactions can reveal regressions. Extracted output is not automatically a correct answer or training dataset. The local rebuild has been checked using synthetic documents; it has not established accuracy on a representative collection of real bank statements.

Document retention and deletion · Review PDF to Excel · Structured CSV input