# How to convert a PDF table to Excel or CSV for free: bank statements, invoices, and why columns shift

> A PDF has no real table inside it, only words placed at positions on a page. Here is how the converter rebuilds rows and columns from those positions, what the three column settings do, which numbers it turns into real numbers (and which it leaves alone), and how to tidy the sheet afterwards.

Published 2026-10-23 by Toolcloak. 7 minute read. Original: https://toolcloak.com/articles/how-to-convert-a-pdf-table-to-excel-or-csv-free-bank-statements-invoices-and-why-columns-shift
Related free tool: [PDF to Excel Converter](https://toolcloak.com/tools/pdf-to-excel)

You have a bank statement, an invoice list or a marks report as a PDF, and you want it in Excel so you can add, sort or filter it. Copying from the PDF and pasting usually gives one messy column. The [PDF to Excel Converter](https://toolcloak.com/tools/pdf-to-excel) does better, because it rebuilds the table from where each word sits on the page. It is not magic, though, and knowing how it thinks tells you why a column sometimes lands in the wrong place and how to fix it in one click.

> **In short**
> - Choose a PDF, pick **Excel (.xlsx)** or **CSV**, and press **Convert to Excel**. Text-based PDFs of up to **100 pages** work.
> - It rebuilds rows and columns from **text positions**. A PDF has no real table inside, so always check the sheet.
> - **Column detection** has three levels: Tight, Normal (the default) and Loose. Change it when columns are split too much or joined together.
> - **Turn numbers into real numbers** is on by default. It strips ₹, $, € and £ and commas, and turns 12% into 0.12. Turn it off for phone, card or account numbers.
> - A **scanned PDF** has no text to read. The tool says so and points you to OCR.

## Why a PDF table is not really a table

A Word or Excel table knows its own rows and columns. A PDF does not. It only stores instructions such as “draw the word “Salary” at this spot, and draw “52,000” at that spot”. The neat grid you see is just words and lines placed in the right positions. So any tool that reads a PDF has to **guess** the table by looking at where the words are.

That guess is a good one when the PDF was made by software from a real table, like most bank statements, GST reports and school mark sheets. It is a weak one when the layout is loose, centred or mixed with pictures.

**How the converter rebuilds a table**

1. **Read every word and its position.** For each piece of text, the tool notes how far across and how far down it sits on the page.
2. **Group words into rows.** Words at almost the same height (within a few points) are treated as one line.
3. **Join nearby words into cells.** Words closer together than the column-detection gap are glued into one cell, so “Opening Balance” stays together.
4. **Line the cells up into columns.** Cells that start at about the same distance from the left edge fall into the same column.
5. **Save as a sheet.** You get an .xlsx file, or a .csv, named after your PDF.

*Rows follow height and columns follow left edges. Most problems come from words that do not line up neatly on those two things.*

## How to convert, step by step

1. Open the [PDF to Excel Converter](https://toolcloak.com/tools/pdf-to-excel) and choose your PDF.
2. Pick **Output**: **Excel (.xlsx)** for a ready spreadsheet, or **CSV** for a plain text file that every program reads.
3. Pick **Pages**: all pages on one sheet, or one sheet per page (Excel only).
4. Leave **Column detection** on Normal for the first try.
5. Leave **Turn numbers into real numbers** on unless your PDF has long identifiers (see below).
6. Press **Convert to Excel**, then download. The note under the result tells you how many rows were found, and reminds you to check the sheet.

Try it: [PDF to Excel Converter](https://toolcloak.com/tools/pdf-to-excel). Choose a text-based PDF, pick Excel or CSV, and convert. Check the first rows before you rely on the whole sheet.

**The converter at a glance**

|  | Value | Note |
| --- | --- | --- |
| Pages at most | 100 | Longer PDFs are refused |
| Output types | 2 | Excel (.xlsx) or CSV |
| Column levels | 3 | Tight, Normal, Loose |
| Uploads | 0 | It runs in your browser |

*If your PDF is longer than 100 pages, split it first with the [Split PDF guide](https://toolcloak.com/articles/how-to-split-a-pdf-into-single-pages-or-equal-parts-and-pull-out-only-the-pages-you-need) or pull out only the pages with the table using [Extract PDF Pages](https://toolcloak.com/articles/how-to-extract-pages-from-a-pdf-keep-only-the-pages-you-need-in-any-order-and-save-them-as-a-new-file).*

## Which Column detection should you choose

This one setting decides how close two words must be to count as one cell. It is the first thing to change when the result looks wrong.

| Setting | Words closer than this stay together | Use it when | Typical problem it fixes |
| --- | --- | --- | --- |
| Tight (many small columns) | 6 points | A neat table with narrow gaps between columns | Two columns glued into one cell, like a date and a time |
| Normal | 12 points | Most statements and reports | Start here |
| Loose (fewer, wider columns) | 22 points | Text that is spaced out or split into many tiny cells | A name or address broken into three columns |

*A point is 1/72 of an inch. The three numbers are small steps, so changing the level slightly changes where cells break. Try the neighbouring setting, not the extreme one.*

> **A quick test:** Convert the first page only if you can (Extract PDF Pages makes this easy). Look at the headings row. If a heading such as “Date” and “Description” sit in the same cell, go Tighter. If one heading is split into two cells, go Looser. Fix it on one page and then run the whole file.

## What “Turn numbers into real numbers” does

Text in a PDF is just text. “52,000” is the characters 5, 2, comma, 0, 0, 0, so Excel would treat it as a word and refuse to add it up. With this option on, the converter changes cells that look like numbers into actual numbers, so SUM, sort and filter work straight away.

| In the PDF | In the sheet | Why |
| --- | --- | --- |
| ₹1,24,500.50 | 124500.5 | Currency sign and commas are removed, Indian grouping included |
| $2,450 or £99.99 | 2450 or 99.99 | Dollar and pound signs are removed too |
| 12% | 0.12 | Percent becomes a fraction. Format the cell as Percentage in Excel to see 12% |
| -350 | -350 | A minus sign is kept |
| 1 200 | 1200 | A space inside a number is removed |
| 007 | 007 (text) | A leading zero means it is a code, so it is left as text |
| 2026-10-23 | text | Dates are left alone, so use Excel’s Text to Columns or date format |
| (1,200) | text | Brackets for negatives are not converted, so replace them yourself |
| 12:30 | text | Times are left alone |

*The cell is changed only when the whole value is a plain number. Anything with letters, extra symbols or a leading zero stays exactly as it was.*

> **Turn it off for long identifiers:** A card number written as “4521 8890 1234” has spaces, so it becomes the single number 452188901234. A phone number such as 9876543210 becomes a number too. Excel keeps only 15 significant digits, and shows long numbers in scientific form, so a 16-digit card or account number will be damaged. If your PDF has phone numbers, account numbers, PAN, Aadhaar or invoice numbers made only of digits, switch this option off and convert them as text.

## Scanned PDF: why you get “No text was found”

There are two very different kinds of PDF. A **text-based** PDF, such as a statement downloaded from a bank’s website, contains real words. A **scanned** PDF is a photograph of paper, so as far as the computer is concerned it has no words at all, only a picture. The converter can only read words, so for a scan it stops with a message that the file looks scanned.

**Which kind of PDF do you have?**

| Text-based PDF | Scanned PDF |
| --- | --- |
| Works well | No text to read |
| You can select and copy a word with your cursor | Selecting text selects nothing, or the whole page |
| Downloaded from a bank, GST portal or office software | Made by a scanner or a phone camera |
| Rows and columns are rebuilt from positions | Needs OCR first, and tables may still need typing |

*A simple test: open the PDF and try to highlight a single word. If you can, it is text-based.*

For a scan, [OCR](https://toolcloak.com/tools/image-to-text-ocr) turns the picture into text (see the [OCR guide](https://toolcloak.com/articles/how-to-copy-text-from-an-image-or-screenshot-ocr-in-english-hindi-and-other-languages) for how). OCR gives you text, not a table, so expect to arrange it in columns yourself. Document pages you photographed are better scanned with the [Document Scanner](https://toolcloak.com/tools/document-scanner) first, so the text is straight and clear. If the PDF is locked with a password, unlock it first with the [Unlock PDF guide](https://toolcloak.com/articles/how-to-remove-a-password-from-a-pdf-when-you-know-it-and-what-to-do-when-you-do-not).

## Why the sheet looks wrong, and the fix

**The four usual problems**

| Box | What it means |
| --- | --- |
| A column drifts or is split in two | Text that is centred or right-aligned starts at a slightly different spot on each row, so it can fall into a neighbouring column. Try the next Column detection level. |
| One record is spread over two rows | A long description that wraps onto a second line becomes its own row, because it sits at a different height. Merge those rows in Excel, or delete the extra row. |
| The heading repeats on every page | With all pages on one sheet, the table header printed at the top of each page is copied each time. Filter the sheet and delete the repeats. |
| Numbers are still text | The option was off, or the value has letters, brackets or a leading zero. Fix it in Excel with Find and Replace or Text to Columns. |

*Every one of these is a result of rebuilding a grid from positions. They are normal, and each takes a minute to fix in Excel.*

## Excel or CSV, one sheet or many

| Choice | Pick it when | Good to know |
| --- | --- | --- |
| Excel (.xlsx) | You will work with the numbers in Excel or Google Sheets | Sheet is named “PDF data”, or “Page 1”, “Page 2” and so on |
| CSV (.csv) | You will import it into another program, a database or a script | Always one block of rows. The file has a marker at the start so Excel reads ₹ and Hindi correctly |
| All pages on one sheet | You want to sort, filter or total the whole report | Header rows from each page come along |
| One sheet per page | Each page is a separate section, such as one class per page | Excel only. CSV is always a single list |

If you are going to feed the data to a program, [CSV to JSON](https://toolcloak.com/articles/csv-to-json-how-to-convert-with-commas-quotes-and-line-breaks-inside-fields-and-the-mistakes-that-corrupt-data) and [JSON to CSV](https://toolcloak.com/articles/how-to-convert-json-to-csv-nested-objects-lists-inside-cells-and-opening-the-file-in-excel-without-broken-columns) explain the format traps that matter. If what you really need is editable text and not a table, [PDF to Word](https://toolcloak.com/articles/pdf-to-word-why-the-layout-breaks-text-pdf-against-scanned-pdf-and-how-to-get-editable-text-out) is the better route.

## A 10-minute cleanup routine in Excel

**From converted file to a sheet you trust**

| When | What |
| --- | --- |
| 1 | Compare the totals. Add up one column in Excel and compare it with the total printed in the PDF. If they match, the numbers came through correctly. |
| 2 | Fix the header. Delete the repeated heading rows and keep one header at the top. |
| 3 | Merge broken rows. Join any record that was split across two rows. |
| 4 | Set the formats. Make dates real dates, percentages real percentages, and IDs text. |
| 5 | Save a copy. Keep the original PDF next to the sheet, in case you need to check a figure. |

*Step 1 is the most important. A matching total is the best proof that nothing was lost or misread.*

> **Privacy:** Bank statements and invoices are sensitive. The conversion runs in your browser and the file is not sent to a server, which is the main reason to prefer a tool like this over an upload site. Still, remove the downloaded files when you are finished.

## Frequently asked questions

### How do I convert a PDF to Excel for free?

Open the PDF to Excel Converter, choose your PDF, pick Excel (.xlsx) as the output and press Convert. It works on text-based PDFs of up to 100 pages and runs in your browser, so the file is not uploaded.

### PDF se Excel kaise banaye?

PDF to Excel Converter kholiye, PDF chuniye, Output mein Excel (.xlsx) rakhiye aur Convert dabaiye. Text wali PDF ke table rows aur columns mein aa jate hain. Download ke baad totals ko original PDF se milakar dekhiye.

### Why does it say no text was found?

Your PDF is probably scanned, meaning it is a picture of paper with no real words in it. The converter can only read text-based PDFs. Run OCR on the pages first, then arrange the text into columns.

### Why are my columns in the wrong place?

PDFs store word positions, not tables, so columns are guessed from where words start. Centred or right-aligned text can drift. Change Column detection to Tight or Loose and convert again.

### Why are some numbers still text in Excel?

Only plain numbers are converted. Values with brackets, letters, dates, times or a leading zero are left as text on purpose. Check that the option is on, and fix the rest with Find and Replace or Text to Columns.

### Can it convert a bank statement PDF to Excel?

Yes, if the statement is text-based, which downloaded statements usually are. Remove repeated headers afterwards, merge any wrapped lines, and compare a column total with the printed total. Turn off number conversion if you want account numbers kept exactly.

### What is the difference between Excel and CSV output?

Excel gives you a spreadsheet ready to open, with one sheet or one sheet per page. CSV is a plain text list that any program can import, and it is always a single block of rows.

### Is my PDF uploaded anywhere?

No. The file is read in your browser and the result is created on your device. That is useful for statements and invoices, but you should still delete downloaded copies you no longer need.