PDF to Excel

Each table's columns are worked out from its own rows, by where their text starts and ends, which is what lets borderless tables come out as a grid; page titles, headings and paragraphs around the table take no part. Values are then parsed into real Excel numbers: thousands separators removed, bracketed negatives such as `(100.00)` turned into -100 while Convert Parentheses is on, and currency symbols stripped when Strip Currency is on. The Bank Statement, Invoice and Expense presets turn Strip Currency on; Custom / None leaves symbols in place, so those amounts stay text. A cell with a second number beside the amount, such as `$10.00 2`, keeps its symbol and stays text rather than becoming a wrong number. Cells merged across several columns in the original are not detected as merges; their text lands in the first column it overlaps, so a stacked or merged header row is worth a quick look before you build formulas on it.

With Auto-Stitch selected, a table that carries on from the page before is added to the same sheet, so a multi-page statement comes out as one continuous worksheet with the repeated header dropped. A page that prints the header again is joined when the column names match; a page that starts straight with data is joined when it has the same columns, or leaves one of them empty, as a page with no deposits does. A page without a header that does not line up stays on a sheet of its own with a warning. The joined sheet is named for its page range, and the Diagnostics tab counts the stitched sheets. Choose Separate Sheets if you would rather keep one sheet per page.

Yes. Excel (.xlsx), comma-separated, tab-separated and structured JSON all run through the same cleaning, so the numbers match whichever you pick. Dates and percentages become real Excel values only in the .xlsx; the other formats keep them as printed. A single CSV or TSV download holds one table, the one on screen, and a ZIP of several files gives each table a file of its own. In CSV and TSV, a cell that starts with `=`, `+`, `-` or `@` and is not a plain number, such as `+91-98765-43210`, gets a leading apostrophe, because a spreadsheet application runs such a cell as a formula on open, and that is a well-known way for a data file to run something you did not intend. In the .xlsx it is a plain text cell.

Check the percentage next to each sheet on the Diagnostics tab. On a page read from its text layer it is the share of the reconstructed grid that actually holds content, so a low number usually means the page was not really a table and the column detection had nothing to lock onto. On a scanned page the same slot shows the recognition engine's character confidence instead, which is a different thing entirely — a low value there means the scan is hard to read. Warnings appear under the sheet name in both cases.