Pull tables out of a PDF into a spreadsheet. Ruled tables are recognised well; a scan with no text layer is reported as such rather than handed back as an empty sheet.
Check that the numbers landed in the right columns.
This one needs an upload, and is deleted afterwards
Browsers cannot do this calculation, so the file goes to our server. It is deleted automatically 60 minutes after the job finishes, and the working copy lives in a memory filesystem that is wiped whenever the server restarts.
We do not log file names and do not log file contents — that is a hard rule in our engineering standards, not a setting.
What it does — and what it doesn't
Ruled tables work best. Tables held together only by whitespace are harder, and a table running across pages has to be joined up by hand.
A scan with no text layer cannot be recognised, and you are told so explicitly instead of receiving an empty result. Run OCR first.
Merged cells are flattened into the top-left cell.
Numbers become real Excel numeric values, and the CSV carries a UTF-8 BOM so Excel on Windows opens non-Latin text correctly.
Common questions
The columns are wrong.
Usually the original has no rules and relies on whitespace. Try CSV and use your spreadsheet's text-to-columns tool.
Can I do just one page?
Split that page off first, then convert.
Why are some complex tables unsupported?
We implemented table detection ourselves — the usual library shells out to a component whose licence we cannot accept. Ours is solid on regular tables and weaker on artistic layouts.
