How to Convert PDF Tables to Excel (and Fix the Ones That Break)
PDF to Excel is the hardest of the common conversions. A PDF does not know what a table is — it only knows where characters sit on a page. Everything else is inference, and inference goes wrong in predictable ways.
Why tables are hard to recover
In Excel a cell is a concept: row 3, column B, containing this value. In a PDF that cell does not exist — there is only a number drawn at a particular x and y coordinate.
The converter has to notice that a group of numbers forms a column. It succeeds when the table came from software and the gaps are consistent. It fails when a cell wrapped onto two lines, when a column is very narrow, or when a row spans a page break.
That is why the same tool produces a perfect spreadsheet from one PDF and a mess from the next. The difference is the source, not the tool.
Convert
- Open PDF to Excel and add your file.
- Choose the page range. Converting only the pages with tables is faster and avoids empty sheets.
- Convert, then look at column A. If the whole table landed in one column, separator detection failed.
- Compare row counts. A 40-row table that produced 80 spreadsheet rows was split at every wrapped cell.
- Fix the first few rows by hand, then copy the pattern down instead of correcting every cell.
Fixing a broken conversion
- Everything in column A: use Data → Text to Columns and pick the visible separator, usually a wide space.
- One row became two: a wrapped cell. Delete the second row and merge the text back.
- Numbers stored as text: select the column, then Data → Text to Columns → Finish. That coerces them.
- Leading zeros lost on account numbers or postcodes: format the column as Text before pasting.
- Merged header rows became separate cells: rebuild the header manually. Faster than automating it.
- A table split across a page break arrived as two tables: concatenate, then delete the repeated header.
When re-typing is genuinely faster
For a table under about twenty rows with clean columns, typing it is often quicker than repairing a bad conversion, and the result is exact.
For a scanned table that is even more true. OCR of tables is the least reliable form of OCR, because it must get both the characters and the geometry right.
Rule of thumb: if repairing takes longer than thirty seconds per row, start fresh.
Privacy
The PDFs people convert to Excel are financial: invoices, payroll runs, bank statements, supplier price lists. That is the complete dataset, not a page of it.
A browser-based converter parses the PDF and builds the workbook on your own device. Nothing is uploaded — which for a payroll table is the only acceptable arrangement.