PDF to Excel

How a table is recovered from a PDF

A PDF does not contain tables. It contains pieces of text drawn at coordinates, and a table is just an arrangement of them that a reader recognises. Recovering it means working that arrangement out again from the positions.

Text is first grouped into rows by its baseline — pieces whose baselines sit within a fraction of the type size of each other are one row. Within a row, a gap wider than most of a letter starts a new cell, while narrower gaps are the spaces between words. Columns are then found across the whole page: cells that overlap horizontally from row to row belong to the same column, which is the same vertical channel of white space your eye follows down a table. A heading that spans several columns is kept out of that vote, so it cannot weld two columns together.

Text running in another direction is set aside, and a page turned sideways is read across rather than down. The grid in the preview is exactly what goes into the file.

Numbers become numbers — but only when there is no doubt

A spreadsheet of numbers stored as text is useless: SUM returns zero and sorting puts 100 before 20. So cells that are unambiguously numbers are written as numeric cells. The rules are strict on purpose. Thousands separators are accepted in consistent groups, western (1,234,567) or Indian lakh and crore grouping (12,34,567). Negatives are recognised as -12, −12, (12) in accounting brackets, and 12- as some accounting systems print them. A currency symbol or code before or after the number — $, €, £, ¥, ₹, Rs., INR, USD — is dropped, since Excel stores the amount and formats the currency separately.

Everything else stays text, because converting it would change what the cell says: identifiers with leading zeros like 00123, percentages, dates, phone numbers, and anything longer than 15 digits, which Excel would round. The decimal point is detected from the document — 1,234.56 or 1.234,56 — and can be set by hand if the guess is wrong. Untick Turn numbers into numeric cells to keep every cell exactly as printed; plain numbers like 42 are still numbers in Excel either way.

Bank statements, reports and price lists

Most people looking for PDF to Excel have a statement. Banks and card issuers send PDF statements, and reconciling one in a spreadsheet beats retyping it. Each page becomes its own sheet by default; choose All pages in one sheet to get one continuous table, with a blank row where each page ends. Statements repeat their column headings on every page, so delete those repeats after importing.

Statement descriptions often wrap onto a second line. Tick Join wrapped lines into the cell above and a line that sits closer than the normal row spacing, with fewer cells than the row above, is joined back into that row. It is a heuristic, which is why it is a choice. Skip lines that are not table rows drops titles, addresses and footnotes that have only one cell, leaving just the table.

Use the CSV download for importing into accounting software or a database; the byte-order mark option is on by default because Excel on Windows needs it to read accented characters correctly. A CSV holds one sheet, so with one sheet per page it contains the sheet shown in the preview.

Limits worth knowing

This reads text-based PDFs — ones exported from accounting software, spreadsheets, websites and word processors. A scanned statement is a picture with no text in it, and the tool says so; run it through OCR first to produce text, which works best on a clean, straight scan.

Merged cells come out in their leftmost column. A table whose columns do not line up — a page of prose beside a chart, or a form laid out in boxes — is not really a table, and the grid will be rough. Very tight columns can be read as one; set the gap option tighter, or wider if words of one cell split apart. Check the preview before relying on the numbers, and compare a total or two against the PDF.