PDF to Excel: The Tool I Almost Didn't Build
Extracting tables from PDFs into spreadsheets is notoriously difficult. I delayed building this one for two months because I was scared of it. Then I did it anyway.
I knew PDF-to-Excel was going to be hard. The way PDF stores data — as positioned text elements on a canvas, with no semantic structure — makes table detection genuinely difficult. There's no tag that says "this is a table." You have to infer it from the positions of text elements relative to each other.
I put it on the to-do list and kept moving to easier tools. Two months passed. It stayed on the list.
Then a user emailed me directly. They were processing quarterly reports — 40+ page PDFs with dozens of data tables — and manually copying the data into Excel. Hours every week. Was this something I could help with?
That email made me stop avoiding it. I spent two weeks on the detection algorithm alone. The core insight was: text elements that share the same Y-coordinate (within a small tolerance) are likely in the same row. Elements with consistent X-coordinates across multiple rows are likely in the same column.
It's not perfect. Complex merged cells still trip it up. Tables with colored backgrounds parse differently than simple bordered ones. But for a standard financial table, government report, or data export, it works remarkably well.
That user replied after I launched it. They said it cut their weekly report processing from 4 hours to 20 minutes. That email is saved in a folder I open when building feels pointless.