Getting a PDF table into Excel with the columns intact

When a spreadsheet is exported to PDF, the table structure is thrown away. What survives is text at coordinates and, sometimes, lines drawn on the page. The grid you can see is the lines; the data is unrelated text that happens to sit between them. Nothing in the file says “this is a column”.

So every PDF-to-Excel converter is doing the same thing: looking at where the text fragments sit and inferring rows from shared vertical positions and columns from shared horizontal ones. It is a reconstruction, and how well it goes depends almost entirely on how regular the original layout was.

• Merged header cells. A header spanning three columns is one piece of text with one position, so it lands in one cell and the two beside it come back empty.
• Values that wrap onto two lines. Two vertical positions means two rows, and a long product description splits itself in half.
• Numbers with spaces as thousand separators. “12 345” is two fragments with a gap between them, and a gap is exactly what a column boundary looks like.
• Blank cells. Nothing is drawn where a value is missing, so there is no evidence a column exists there at all, and everything to the right can shift left by one.
• Right-aligned numbers next to left-aligned text. The two columns are detected at different positions and can be read as three.

1. 1Confirm the PDF has real textTry to select a number in the table. If you cannot, this is a scan and no converter will extract it without OCR — a different job with a different accuracy problem.
2. 2Convert, then check the column count before anything elseNot the values — the count. If your six-column table came out with eight columns, a gap was misread and everything downstream is shifted. That is worth catching in five seconds rather than after you have started formatting.
3. 3Check the header row and the last rowHeaders break because of merged cells; the final row breaks because of totals in a different layout. Both ends are where the assumptions fail.
4. 4Spot-check three numbers against the PDFPick one from the top, middle and bottom. A column shift shows up immediately, and a shifted column is the failure mode that quietly survives into a report.

Conversion is worth it for a table you cannot get any other way. It is not always the fastest route.

• For a table of twenty rows, retyping it is often quicker than fixing a bad extraction — and you will read the numbers as you go, which catches errors a conversion hides.
• If the PDF came from a colleague, ask for the spreadsheet. Ten seconds of asking beats an afternoon of repair.
• If it is a recurring report, the source system almost certainly has a CSV export. Somebody chose PDF; that choice can usually be changed.
• If it is a scan of a printout, and accuracy matters, treat OCR output as a draft to be checked rather than data to be trusted.