When your Gorenjska Banka 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. When a statement document has been scanned or photographed rather than exported as a clean digital file, the engine applies OCR to read the transaction table from the image. It then runs the same verification it would on a machine-generated page: the printed running balance is checked against each row, line by line. A photograph with uneven lighting, a slight skew, or a shadow across a column can still produce a reliable extract because the balance check catches any digit that the OCR misread. The engine does not reject the whole file when one entry fails; it flags the suspect row so that you can review and correct a single figure instead of re-entering the entire month.
If a transaction description on your Gorenjska Banka statement is long enough to wrap across two or three lines, it still needs to land in a single row once converted. Transaction descriptions that run long and wrap across multiple lines are a common feature of statement documents, and a frequent source of misalignment in manual copy-and-paste workflows. The engine detects multi-line entries by analysing row geometry and consistent column alignment, not by relying on every transaction occupying exactly one line. A wrapped memo line that breaks across two or three rows is reassembled into a single record before extraction, so a description split over several lines ends up in the correct cell rather than being fragmented across separate rows in the output.
However your Gorenjska Banka statement lays out its columns, that layout is only useful to you once it is mapped correctly into the export. When you map each column from the extracted table into the right place in the export, you decide how the final file is structured. The engine presents the columns it found in the statement, and you assign each one to the appropriate field in the target format, whether that is a single-sheet Excel workbook, a CSV, or a QBO file. Running-balance and date columns can be included, reordered, or omitted to match the layout your ledger or import routine expects. The mapping is saved so that the next statement from the same layout lands with the same columns already aligned.