Why PDF tables need checking after they come out
Copying a table out of a PDF usually gives you one long column of words, because a PDF does not store tables at all: it stores words and where to draw them. Getting a spreadsheet back means working out the rows and columns again from those positions, which works well on neat tables and needs checking on messy ones.
Updated
Which tool you need
A bank or card statement
Choose Every line, in columns. A statement page is one long table, and reading every line keeps the transactions a table finder might pass over.
A report with tables between paragraphs
Keep Find the tables, which leaves the paragraphs out and returns only text that lines up in columns.
Several tables you want in separate sheets
Turn on One file per table to get a ZIP with a CSV for each, ready to open one by one.
The table has to go back into a PDF after editing
CSV to PDF sets a CSV as a paged table with its header on every page.
What to check before trusting the numbers
Look at the columns first. A cell whose text wraps onto a second line can turn into an extra row, and a heading that spans two columns can push its neighbors along by one. Both are quick to spot in a spreadsheet and quick to fix there.
Then check a total. If a statement's balance or a report's sum still adds up after conversion, the figures came through; if not, the row that went astray is usually near a page break or beside a long description.
What the CSV holds
Values separated by commas, one row to a line, every row padded to the same number of columns. A table that carries on over a page break with the same columns is joined into one, and a header row repeated at the top of the next page is dropped rather than appearing twice.
A cell that starts with =, +, - or @ is written with an apostrophe in front, so a spreadsheet shows it as text instead of running it as a formula. Ordinary negative amounts such as -12.50 are left alone, because they are figures, not formulas.
Which tables come out poorly
The lines drawn around cells are never read. Columns come only from the white space between words, so a table whose columns are divided by ruled lines with hardly any space either side can run its columns together.
Every line, in columns turns every line of every chosen page into a row, including a cover page's title or a page of terms. Choosing only the pages the table is on saves deleting those rows afterwards. Find the tables, on pages where nothing lines up, says no table was found rather than writing an empty file.
What to know before you start
- Scanned tables are pictures and have no words to read.
- Merged and multi-line cells are the usual source of misplaced values.
- Colors, borders and formulas have no place in a CSV.
- Text set sideways, such as a rotated column heading, is left out of the CSV.