PDF to Excel

How it works

1. Drop the PDF. Bank statements, invoices, price lists and reports are what this is for.
2. Choose a sheet per page, which keeps a long statement readable, or one sheet for the whole document when the same table runs across several pages.
3. Leave “Turn figures into numbers” on unless the column holds things like order references, which should stay exactly as printed.
4. Press Convert to Excel and download the .xlsx. It opens in Excel, Numbers, LibreOffice and Google Sheets.

Frequently asked questions

Is my PDF uploaded?

No. The PDFs people want as spreadsheets are bank statements, invoices and payroll reports — the documents you would least like to leave on someone's server. This reads the file in your browser and writes the workbook there too.

How does it know where the columns are?

The same way you do when you look at the page. A PDF holds no table, only glyphs at coordinates, so the page is projected onto its horizontal axis and every strip that no row writes into becomes a column boundary. Rows come from the pieces of text that share a baseline. It works well on the kind of table that is printed with real gutters between the columns, which is most of them.

When does it get it wrong?

Three cases, and they are worth knowing before you trust the output. A cell that spans two columns pulls its neighbours together. Columns set so tightly that no clean gap runs down the whole page are read as one. And a table where the columns are marked by ruled lines but the text is spaced evenly across them can split in the wrong place. Check the first few rows against the PDF; if the shape is wrong, the answer is usually to convert one page at a time.

What happens to ordinary paragraphs?

They come out as one column, a row per line, which is the honest answer: prose is not a table, and inventing columns for it would be worse than leaving it alone.

Which cells become numbers?

Anything that is plainly a figure: 1,250 and $1,250.75 come across as numbers, thousands separators and the currency symbol dropped, and a figure in brackets becomes a negative, as an accountant writes it. A percentage, a date and anything with letters in it stay as text, where they still mean what they say. Switch the option off to keep every cell exactly as printed.

My PDF is a scan and nothing came out.

Put it in as it is. A scan is a picture of a table with no text in the file, so the page is read with the same recogniser the OCR tools use, and the boxes it finds — one per cell, in a table — go through the same column reasoning as any other page. The first scanned file waits for the model to download; after that it is quick. What it cannot rescue is a photograph taken at an angle or in poor light: straighten and light the page first and the difference is large.

What are the free limits?

Three files a run at 20 MB each and 10 runs a day. The workbook itself carries no watermark or footer, on any plan.