PDF to Excel

Every cell is written to the workbook as a string, which is why Excel flags columns of figures with its "number stored as text" marker. The reason is fidelity: a value on a PDF page might be 1.234,56 in a European layout, (89.00) in accounting notation, or 12,5% with a footnote marker attached, and guessing at a conversion would silently change your data. Converting after the fact is easy and reversible — select the column in Excel and use Convert to Number, or run a VALUE() pass in any spreadsheet — while unpicking a bad automatic conversion is not. Dates behave the same way and need a text-to-columns pass with the right locale before they will sort chronologically.

Pick one sheet per page when a page's tables belong together — a continued table split by a page break, or a main table with a footnote table underneath — because they are stacked into a single worksheet with a blank row between them and stay in reading order. Pick one sheet per table when the tables are independent and you intend to sort, filter or pivot each one, since a worksheet holding two stacked grids of different widths is awkward to work with. The preview tabs show exactly what each layout produces, and switching between them is instant because the document is not re-read. Pages with no detected table are skipped under both layouts.

Sheet names are generated from the source: "Page 7" in the per-page layout, "Page 7 Table 2" in the per-table layout. Excel imposes rules the tool has to satisfy, so names are clipped to 31 characters, and the characters Excel forbids in a sheet name — backslash, forward slash, question mark, asterisk, square brackets and colon — are replaced with spaces. If clipping would produce two identical names, a numeric suffix is appended so the workbook stays valid. Nothing is taken from the table's own heading, because a detected heading is unreliable and frequently sits inside the first data row rather than outside the grid. Rename the tabs in Excel once the file is open.

There is no page filter in PDF to Excel — it converts every page where it detects a table, and the only control is how those tables are distributed across worksheets. If you need a subset, either delete the unwanted sheets after opening the workbook, which takes seconds in the per-page layout, or isolate the pages first with Extract PDF pages and run the smaller PDF through here. PDF to CSV does have a page-range field, so if narrowing the source pages is the main thing you need and one flat file is acceptable, that tool is the shorter route. Both use exactly the same detection, so whatever one finds the other finds too.

The table detection behind PDF to Excel is accurate in proportion to how cleanly the PDF was produced. Tables generated by a reporting tool or exported from a spreadsheet, with consistent column positions and clear gaps between columns, come through cleanly. Financial statements with indented sub-totals, merged header bands, or figures right-aligned into columns that nearly touch are where it struggles, because columns are inferred from horizontal gaps in the text rather than from ruling lines the page draws. Scanned pages produce nothing at all, since there are no text runs to cluster — run OCR PDF first. Always click through the sheet tabs before downloading; the ten-row preview per sheet exists so a misread grid is caught before it reaches your model.

No part of the workbook build touches a server. The PDF is read locally, detection runs in the tab, and SheetJS assembles the .xlsx bytes in your browser before handing them to a download — which is why a spreadsheet full of salary, invoice or patient data can go through this tool without leaving your machine. There is no account, no conversion quota, and nothing is added to the workbook: no branding sheet, no footer row, no metadata beyond what the spreadsheet format itself requires. The workbook builder is loaded as a separate code chunk the first time you press Download, so that first export fetches a script; your document is not part of that request.