PDF to Excel

How the columns are found

The converter reads the position and width of every run of text on each page, then decides which values belong in the same column by where they fall horizontally across all the rows of that page.

• Text on the same baseline forms a row. Words close together are kept as one cell, so "New York" stays in a single cell rather than splitting in two.
• A gap wider than about one character is treated as a cell boundary, which is how most statements and reports space their columns.
• Values that overlap horizontally on any row are placed in the same column. That is what keeps right-aligned amounts lined up under their heading even when every number is a different width.
• A row with a missing value keeps an empty cell in that position, so the values to its right do not slide into the wrong column.
• A title or a heading that runs across the page is placed in the first column, rather than being forced into one of the table's columns.
• Rows where every value is bold are kept as bold text in Excel, which is usually the header row.

Numbers Excel can actually use

• The most common reason a pasted PDF table will not add up is that the numbers arrived as text. With the numbers option on, values are converted into real numbers that sum, sort and chart correctly.
• Thousands separators (1,234.50), currency symbols ($, €, £, ¥), accounting negatives in brackets such as (1,200), trailing minus signs and percentages are all recognised. A percentage is stored as its true value, so 12.5% becomes 0.125 and displays as a percentage.
• Anything that would be risky to guess stays as text. Values with leading zeros, such as account numbers, postcodes and invoice IDs, keep their zeros. Dates stay exactly as printed. The European format 1.234,56 is left as text rather than misread as one thousand and something.
• Text that happens to begin with an equals sign is written as plain text, never as a formula, so a malicious PDF cannot plant a live formula in your spreadsheet. The CSV option neutralises the same pattern for apps that read CSV.
• Header rows are kept as text, so a column heading such as 2026 stays a label rather than turning into a number.

Where it needs a hand

• A scanned statement is a picture, so there is nothing to read, and the tool will say so. Run it through OCR first, then convert the result. Check every figure afterwards, because recognition can misread 0 and O, or 1 and l.
• Text that wraps onto two lines inside one cell arrives as two rows. Join them by hand, or fix the source if you can.
• Merged header cells that span several columns land in the first of those columns.
• Each page is read on its own, so a table that runs across pages repeats its header row on each sheet. On the single-sheet layout, filter those repeated headers out.
• Two tables side by side on one page are read as one wide table.
• Confirm totals before relying on the figures. If the table has a total row, check the converted values still add up to it. That one check catches nearly every column that has shifted.