When your Lazard statement arrives as a scan or a photo of the page rather than a clean export, the numbers on it are just pixels until they are extracted. A statement that was printed and then scanned or photographed starts as a picture, not a spreadsheet. The engine reads that picture with OCR, extracts the transaction table, and then verifies every running balance against the statement's own printed closing figure. When a smudge or a fold confuses a digit, the verification catches it: the row is flagged, not the whole file. You review one suspect amount instead of rekeying every line.
If a transaction description on your Lazard statement is long enough to wrap across two or three lines, it still needs to land in a single row once converted. Long transaction descriptions that wrap across several lines in a statement document cause ordinary converters to split one entry into multiple broken rows. Convert·Into recognises multi‑line text blocks as a single transaction and maps the full description into one cell in the export, so the narration stays whole. Columns stay aligned; the running balance check still runs across every row, including the ones that spanned three or four lines in the original document.