To convert a scanned PDF to Excel, fix the page images first, run the file through an OCR converter, then check the rows against the original before you use them. A scanned PDF is a picture of a page, so the converter must read characters from pixels, and some will be wrong.
Should you straighten a scanned PDF before conversion?
Yes, when the page is visibly tilted, rotated or surrounded by dark scanner edges. Straightening and cropping help because OCR builds columns from alignment, and a tilted page makes a column drift sideways. A page that already sits square needs nothing.
What is worth fixing first
- Rotation. A page scanned sideways or upside down is the first thing to fix. Many converters detect this, but do not count on it.
- Skew. A tilt of even a degree or two lets the right edge of a column wander over a long page. Re-scan the page flat rather than editing it if you can.
- Dark borders. The black band at the edge of a scan adds nothing and can be read as ruled lines. Crop it off.
- Extra pages. A blank back side, a cover sheet or a bank's advert leaflet adds time and clutter. Remove them before conversion.
What not to do
Do not sharpen, stretch or re-compress the pages in an image editor to "improve" them. Each save as a lossy JPG can soften the digits further, and heavy sharpening can fill in the loops of 6, 8 and 9. If a page is poor, a fresh scan beats editing. The scanner settings that help conversion say what to change on the second attempt.
Before any of this, make sure the file really is a scan. The test for whether a PDF is an image takes ten seconds, and a PDF that does have text needs no OCR at all.
Running the scanned PDF through an OCR converter
An OCR converter reads each page image, recognizes the characters and places them into rows and columns, then writes an Excel or CSV file. Few settings matter, and most of them are choices you make before you click convert, not options on a screen.
Settings that matter
- Language. Pick the language of the print. An engine set to the wrong language misreads accented letters and can mishandle number formats.
- Output format. Choose Excel (XLSX) if you will sort and sum, or CSV if another program will import the file.
- One table or one sheet per page. For a statement you want a single continuous list. If each page lands on its own sheet, merging a PDF into one Excel sheet is covered separately.
- Page range. Convert only the pages you need. A fifteen-page scan with two pages of terms and conditions should not carry them into your sheet.
How the engine decides on columns
You do not choose columns, the engine finds them from alignment. How OCR builds table columns from recognized words explains the mechanism, and why a debit with an empty credit cell is the usual place for a shift.
The seven steps, file to finished sheet
This is the whole job in order. The file names and figures are examples.
- Confirm it is a scan. Open the PDF and try to select a word. If nothing highlights, the page is a picture.
- Check the page count and order. Page 3 should not come before page 2. A scan file named
statement-march.pdfwith pages in the wrong order will give a broken running balance. - Straighten, crop and remove extra pages. Use the fixes above, or scan again.
- Convert with an OCR converter. Choose the language, the output format and the page range.
- Open the result and look at the shape. Are there the right number of columns? Is the date column all dates? Do the amounts right-align as numbers?
- Run the spot-check routine below.
- Fix, then save a clean copy. Correct every misread cell, convert text to numbers where needed, and keep the original PDF next to the sheet.
Checking scanned PDF output against the page
Check scanned PDF output by comparing a few well-chosen rows with the paper, then proving the totals. You cannot read every cell, so choose rows that would hurt most if wrong and use arithmetic to catch the rest.
The spot-check routine
Do these three checks every time:
- Three random rows. Pick one near the top, one in the middle and one near the bottom of the file. Compare date, description, amount and balance with the page. If all three match, the engine read this scan well. If one fails, assume others will.
- Every row over a set amount. Choose a threshold that fits your books, such as every amount over 1,000.00 (an example figure). Large amounts carry the most risk, and a short list is fast to check.
- The closing balance. Opening balance, plus all credits, minus all debits, must equal the closing balance printed on the last page. In Excel that is one formula over the amount columns.
Mistakes that waste the most time
- Converting before checking page order. A swapped pair of pages breaks the running balance and looks like dozens of OCR errors. Fix the order first.
- Trusting a clean-looking sheet. A sheet with neat columns can still hold a wrong digit. Neatness says the layout was found, not that the figures are right.
- Checking only the totals. Two errors can cancel each other out. Match the random rows as well as the closing balance.
- Overwriting the only copy. Keep the original scan until the converted sheet has passed every check, because you may need to look at the page again.
- Fixing cells one at a time without a note. Write down which rows you changed, so a reviewer can see what was corrected by hand.
A worked example
Take a scanned two-page statement with these example figures:
| Check | Page says | Sheet says | Result |
|---|---|---|---|
| Opening balance | 4,200.00 | 4,200.00 | Match |
| Total credits | 6,350.00 | 6,350.00 | Match |
| Total debits | 5,118.40 | 5,181.40 | Differs by 63.00 |
| Closing balance | 5,431.60 | 5,368.60 | Differs by 63.00 |
The two differences are equal, so one debit was misread by exactly 63.00. A transposed pair, 5,118.40 read as 5,181.40, produces that kind of gap. Sort the debits, look for the amounts whose digits could swap, and compare those with the page. You find the fault in minutes instead of reading the whole file.
Excel cannot sum a cell that holds text. A column total that comes out too low can mean cells like "4l2.80" were read as text, so confirm that the number columns hold numbers before you trust any total.
Can Excel convert a scanned PDF by itself?
Not through its PDF import. Excel's Get Data from PDF reads the characters stored in a PDF, and a scanned PDF stores none, so the import finds no table to offer or returns empty cells. Excel does have a separate picture import, which reads an image instead.
The route that does work inside Excel
Excel for Microsoft 365 has Data From Picture, described in Microsoft's Insert data from picture help page. It reads a picture file or a screenshot. To use it on a scan, you would save a page as an image first, then import it page by page. Converting an image to text in Excel walks through that tool, including where it stops being enough.
For a short, clean table that is fine. For a scan of several pages, working page by page invites mistakes at the seams between pages, and a converter built to take the whole file is easier to check. A photo or screenshot that is not a PDF in the first place is covered in image to Excel.
Where the product fits
FastBankConvert takes bank statement PDFs, and its FAQ notes that scanned paper statements take longer than digital ones. The product is getting ready to open, so use the get started page to be told when trial conversions are available. Until then, follow the checks above with whichever OCR converter you use.
Scanned PDF pages with stamps, skew or fold lines
Some pages are better scanned again than converted. A page needs a re-scan when the damage touches the figures: stamps over amounts, a fold line through a column, a skew that no cropping removes, or text that is cut off at the edge.
Which defects you can live with
| Defect | Keep the page? | What to do |
|---|---|---|
| Light skew, figures all readable | Yes | Straighten or accept, then spot-check |
| Fold line between rows | Usually | Check the rows either side of the fold |
| Fold line through a column of figures | No | Flatten the page and scan again |
| Stamp over description text | Often | Retype the description by hand |
| Stamp over an amount | No | Scan again, or read it off the paper |
| Cut-off edge or missing column | No | Scan again with the full page in frame |
| Shadow from a bound spine | No | Press flat on glass or scan a loose sheet |
Rules for deciding
Ask whether a human could still read every figure on the page without guessing. If you must guess, so will the software. Re-scan before you convert, because a misread you caught in the paper is a lot cheaper than one you find after filing.
For a bank statement where the pages are not clean, there is also the simplest fix of all. Ask the bank for a fresh copy, since a digital PDF from the bank avoids the OCR step entirely. And when output looks right but you are changing the layout, keeping formatting after conversion is a separate set of fixes that applies once your numbers are correct.