If a transaction description on your Turkish Bank statement is long enough to wrap across two or three lines, it still needs to land in a single row once converted. Sometimes a transaction description runs long: a reference code, a payee name typed in full, a machine-generated memo that wraps across several lines inside a single row. When the description breaks across lines in the statement document, a naive extraction might split the line mid-field and treat the wrapped text as a new, empty transaction. The engine reads the full row boundary, not just the visible line breaks, so the complete description stays together in one cell. The exported table reflects the actual transaction data without phantom rows or fragmented text.
If your Turkish Bank statement runs to several pages, uploading it starts the same layout detection Convert·Into applies to any statement document, page by page. A statement that runs several pages long can hide things: a missing page in the scan, a page fed upside down, or a page-count mismatch between the header and the actual file. The engine processes every page in the document and cross-checks the transaction stream end-to-end. If a page is unreadable or out of sequence, the resulting gap in the running balance is flagged; a short statement and a long one get the same level of scrutiny, page by page.
Whatever the running balance looks like on your Turkish Bank statement, Convert·Into checks it against the statement's own printed closing figure before you export anything. Every statement carries its own proof: a printed closing balance. The engine verifies the running balance against that figure line by line, from the first transaction to the last. If a digit was misread, a row skipped, or a sign inverted, the final balance will not tie out. The row that breaks the calculation is highlighted, not hidden. The rest of the file remains intact for export, so you review one discrepancy rather than recheck every transaction.