PrivatePDF

Convert a PDF table to Excel

Getting a table out of a PDF is only useful if the numbers land in the right cells. This reads the table's own ruling lines to find its columns, writes one sheet per table, and stores plain numbers as numbers so a total column adds up. Where a table has no ruling lines, it says so rather than guessing — a spreadsheet that opens cleanly and holds the wrong figures is worse than no spreadsheet, because nobody re-checks a file that opened cleanly.

or drop a file anywhere on this panel

opened locally · 0 bytes sent

How it works

  1. 01Open the PDF.
  2. 02Read the panel — it lists every page it found a grid on, before you download anything.
  3. 03Choose whether the first row is a header, and whether to record which page each row came from.
  4. 04Download the .xlsx and check it against the document.

Ruled tables only, and that is the point

A table with visible ruling lines has written down where its columns are: the lines are in the file as exact coordinates, so reading them is precise rather than approximate. A table laid out with whitespace has not written anything down, and recovering it is a research problem whose characteristic failure is plausible-looking wrong data. So this finds ruled tables and tells you plainly when it finds none, instead of returning something that looks like an answer.

Numbers that are numbers, and text that stays text

A cell containing 42 is written as the number 42, so you can sum the column without cleaning it first. A cell containing 1,234 is not, because the thousands separator is read differently in different places and guessing changes the value. Neither is 007, where the leading zeros are significant, nor 12% or £5, where the unit is part of the data. The rule is that nothing is converted unless converting it cannot change what it means.

One sheet per table, named for its page

A long report with a table on eleven pages gives you eleven sheets, each named for the page it came from, rather than everything concatenated into one grid or eleven separate CSV downloads. Header rows are bold and frozen, and columns are widened to fit their contents — so the workbook is usable the moment it opens instead of after ten minutes of formatting.

What it does not understand

Merged cells, and a table that continues across a page break, are not modelled: the first comes out with the value in one of the cells it spanned, and the second as two separate sheets. Both are visible immediately if you compare the sheet with the page, which is why the panel asks you to. A scan has no ruling lines to read — only a picture of them — so there is nothing to find until it has been through OCR.

Questions

Why does it say it found no tables when I can see one?
Almost always because the table is laid out with spacing rather than ruling lines, so there is nothing in the file that says where its columns are. The other common reason is that the page is a scan, where the lines are part of an image rather than real geometry.
Can I get a CSV instead?
Yes — the table-to-CSV tool produces the same extraction as plain CSV files. Use the workbook when you want several tables in one file with formatting; use CSV when something else is going to read it.
Will the formatting from the PDF come across?
No. You get the values, a bold header row and sensible column widths. Colours, borders and fonts from the original are not reproduced, because they are presentation rather than data.
Is the document uploaded?
No. The tables are found and the workbook written by your own browser, and you can confirm it by disconnecting from the network first.