PDF to Excel: Why Tables Come Out Wrong and How to Fix Them

Convert with PDF to Excel first if you have not yet. It runs in your browser, so the PDF — often a bank statement, an invoice or a payroll report — is never uploaded.

Why PDF tables are hard to convert

When a report is saved as PDF, the table stops being a table. What the file keeps is a list of text fragments, each with a position, and a separate set of lines that happen to be drawn between them. Nothing records that a row exists, or that one number belongs under a particular heading.

So a converter measures instead of reading. Text that shares a baseline becomes a row. Text whose horizontal span lines up across several rows becomes a column. Ruled lines, where the table has them, tell a wrapped cell from a new row. That works well on a clean, regular table and less well where the page is irregular — which is where the problems below come from.

How well it works is something we measure rather than claim. On the 158 tables of the public ICDAR 2013 table benchmark — real reports from EU and US government sites — about 93% of the relationships between neighbouring cells come out right. The rest is what this guide is about.

6 common problems and how to fix them

1. Everything landed in one column

Why: The converter found no column positions that repeat from row to row. That happens on a page that is mostly text, on a table turned sideways on the page, and on a table so irregular that its columns cannot be told apart from indents.

How Convertube handles it: A page only becomes a grid when at least two column positions recur across several rows. Otherwise it comes back as one column of lines, on purpose: inventing columns from an indented paragraph would produce something that looks like a table and is wrong. Lines of prose on the same page as a table — numbered paragraphs, notes, captions — are kept out of the column count, so they no longer drag the table below them into a single column.

Fix it in Excel: Select the column, then Data → Text to Columns. Choose "Delimited" if the values are separated by spaces or tabs, or "Fixed width" and click where each column starts.

2. One row became two

Why: A cell whose text wrapped onto a second line in the PDF. The second line has its own baseline, so on its own it reads as a new row.

How Convertube handles it: It joins the second line back into the cell when the page proves it is a wrap: in a ruled table, no line is drawn between the two; in an unruled one, the two lines sit closer together than the table's rows do; and the first line runs right up to the edge of its column. When it cannot tell — a tightly set table with no rules, where a section heading and a wrapped line look the same — it leaves the extra row rather than risk gluing two real rows together.

Fix it in Excel: Join the two cells with a formula such as =A2&" "&A3, copy the result back with Paste Special → Values, and delete the extra row. For a long table, sort by a column that is empty on the broken rows to gather them together first.

3. Numbers will not add up

Why: The cell holds text that looks like a number. A value written as 1,234.50, $48, 12% or (45) carries formatting that a plain number does not, and those stay as text.

How Convertube handles it: A cell that holds a plain number — 1234.5, -12, 48.50 — is stored as a real number, so SUM works on it straight away, and its decimal places are kept as they were written. Anything where the formatting is itself information stays text: leading zeros, thousands separators, currency, dates, phone numbers. Turning an account number like 00417 into 417 would quietly destroy it.

Fix it in Excel: Select the column and run Data → Text to Columns → Finish. Excel re-reads every cell, and values like 1,234.50, $48 or (45) become numbers. If the PDF uses a comma for decimals (1.234,50), click Advanced in the last step and set the decimal and thousands separators before finishing.

4. A value moved into the wrong column

Why: A merged cell — a heading that spans two columns, or a total that sits across several — has no single column of its own.

How Convertube handles it: It places a merged cell in the first column it covers, so it is never dropped and always in the same predictable spot.

Fix it in Excel: Move the cell with cut and paste, or select the columns it should span and use Merge & Center.

5. The table is split over several sheets

Why: Each page of the PDF becomes its own sheet. A long table that runs across pages therefore arrives in pieces, usually with its header row repeated on each.

How Convertube handles it: One sheet per page is deliberate: a page break is often a real table break, and two sheets are far easier to join than one sheet is to take apart once header rows are buried in its middle.

Fix it in Excel: Copy each sheet's rows, without the header, under the first sheet's table. For many pages, Power Query (Data → Get Data) can append the sheets into one table in a few clicks.

6. Empty cells where the PDF meant "same as above"

Why: Many reports leave a repeated value blank — the region is written once and the rows below it are empty. The PDF really is empty there, so the cell arrives empty.

Fix it in Excel: Select the column, press F5 → Special → Blanks, type = and the up arrow, then press Ctrl+Enter. Every blank fills with the value above it. Paste Special → Values afterwards to fix them in place.

Scanned PDFs are a different problem

A scanned PDF is a picture of a page; there is no text to measure. Convertube says so rather than guessing, because text recognition finds lines of text, not cells, and often merges neighbouring cells in ways that are hard to spot. You can run OCR PDF first to make the text selectable, but a scanned table is the one case to check by hand, whatever tool you use.

What you get

Check these three things after converting

1. Pick a total row and SUM the column above it — if the result matches, the numbers came through as numbers
2. Scroll for rows that hold only a fragment of text — a wrapped cell that was left as its own row (problem 2)
3. Compare the row count with the PDF on one page — a missing or extra row shows up at once