How to Turn a Scanned Invoice PDF into Excel
A scanned bill is a photo in a PDF coat. There is no text in the file for a tool to copy. So how do you turn a scanned invoice PDF to Excel? The answer is OCR: tech that looks at the photo and reads the words and sums.
Our scanned invoice to Excel converter runs OCR when a PDF has no text. Then it builds the same clean sheet as it does for text PDFs. This guide tells you what to expect — the good and the weak parts.
Scans are a big share of real bills. Vendors print and sign, sites fax, and drivers snap photos of paper at the end of a run. If your pile has scans in it, you are not the odd one — you are the norm. The good news is that the path from scan to sheet is now a plain one, as long as you know where it is strong and where it needs your eyes.
How OCR reads a scan
First, the tool turns each page into a clear image. Then the OCR reads the image and finds words and numbers. It notes where each word sits on the page. Our tool then groups words into lines, just like it does with text PDFs. From those lines it finds the vendor, the bill number, the rows, and the totals.
A sharp, straight scan reads best. A dark or bent photo reads worse. If you can, scan flat, with good light, at a fine setting. It takes ten extra seconds and saves a lot of fixes.
Where each word sits counts as much as the word itself. A sum at the right edge of a line is a row total. The same sum at the foot of the page is the bill total. OCR hands the tool both the words and their places, and the tool reads the bill the way you do: head first, rows next, totals last. That is why a page that is bent or cut off hurts twice — the words blur, and their places shift too.
Make a scan that reads well
Most bad reads start at the scan, not in the tool. A few habits at scan time fix most of them:
- Scan flat and straight. A page at an angle bends each line, and bent lines break rows.
- Use a fine setting — 300 dots per inch is the sweet spot for print this size.
- Keep the light even. A shadow across the sums is a sum the tool can not read.
- Get the whole page in the frame. A cut-off edge can hold the total.
- Wipe the glass. Dust and smears read as dots and marks on the page.
- Scan in black and white or gray, not a phone photo, when you can. Flat scans beat photos.
You can not always re-scan. Bills come in as they come in — a photo from a site tech, a fax of a fax. The tool does its best with what it gets. But for bills you scan in your own office, this list is the cheapest accuracy you will ever buy: ten seconds at the scanner saves ten minutes at the screen.
What reads well — and what does not
We test our OCR on real bills. On a paired test of 100 bills, the totals came out right 100 times out of 100. Bill numbers came out right 94 times out of 100. Those are the two sums that matter most, and they hold up well.
| Part of the bill | How OCR does on scans |
|---|---|
| Full total | Reads well — right in all 100 test bills |
| Bill number | Reads well — right in 94 of 100 test bills |
| Line items | Weaker — rows can drop or blur |
| Handwritten notes | Poor — check these by eye |
Rows are the weak part. On dense bills, a row can drop or a digit can blur. So for scans, check the rows with care before you export. The totals check helps: if the rows do not add up to the subtotal, you know a row needs a look. More scan facts are in our FAQ.
That totals check is your safety net on every scan. If a row drops, the row sum falls short of the subtotal, and the gap tells you a row is gone. If a digit blurs, the gap tells you where to look. To learn how to read a gap and hunt its cause, see why invoice totals don't add up — the checks there are the same ones to run on a scan's sheet.
Why do rows drop on dense bills? A dense bill packs many thin lines close by each other. On a scan, two thin lines can merge into one in the image, or a faint line can fade out. The head of the bill — big print, wide space — stays clear, which is why the number and total hold up while a mid-table row can slip. If your scans are dense, give the middle of the table your closest look. The top and foot of the bill will take care of themselves.
Text PDF or scan? How to tell
Open the file and try to select a word with your mouse. If you can highlight the word, the PDF has text. If the whole page moves as one photo, it is a scan. Our tool can tell on its own and picks the right path. You do not have to set a thing.
Thick scans take more time. A dense scan with many pages can take a few minutes to read. That is normal. Short text PDFs are done in seconds. Either way, the sheet you get is the same shape, so your checks and exports work the same. If most of your bills are scans with many rows, read how to get line items out of an invoice PDF for the checks that matter.
Mixed piles: scans and text bills together
Real piles are mixed. In one month you may get text PDFs from your big vendors, scans from small ones, and photos from the field. You do not have to sort them by kind first. The tool tests each file on its own: if it finds text, it reads the text; if it finds none, it runs OCR. The sheet at the end is one shape for all of them, so your sums, sorts, and exports work the same across the whole pile.
Spend your check time where the risk is. Text bills read near clean; a fast scan of the flagged bills is most often enough. Scans earn a closer look at the rows, as this guide has said. That split — light check for text, close check for scans — is the honest way to work a mixed pile. It puts your eyes on the bills that can hurt you and takes them off the ones that can not.
One habit helps the split: name scan files with a mark, like “scan-” at the start of the file name, as they come in. Then the pile sorts itself, and you know at a glance which sheets came from OCR and need the row check. Small marks, big calm.
When to re-scan, and when to type
Some scans should not be read twice — they should be made again. If the total comes out wrong, or the sheet is full of gaps, the scan is too poor to trust. A fresh scan, flat and sharp, will read better than any fix you make by hand. Re-scanning takes a minute. Fixing twenty blurred rows takes far more.
Some parts should just be typed. Handwritten notes — a site note, a signed “received” — read poorly no matter what. If a bill's key facts are by hand, type those few fields and let OCR do the printed rest. And the best fix of all: if the vendor sent a scan of a bill they made on a computer, ask them for the text PDF. The text was in their file all along. Many vendors will send it with no fuss, and a text PDF reads clean every time.
A rule of thumb sorts the two in seconds. If the bill is short and the scan is poor, type it: five fields by hand beat twenty fixes. If the bill is long and the scan is fair, run OCR and check the rows: the read will beat your typing on time, and the totals check will tell you if the read was sound. Poor scan plus long bill is the one hard case — that is the bill to re-scan, since both paths are slow on it.
Quick questions
Will it work on a photo from a phone? Yes, if the photo is sharp, flat, and well lit. Lay the bill on a table, hold the phone straight above it, and fill the frame with the page. A photo shot at an angle from a desk chair is the top cause of poor reads we see.
Does OCR change my file? No. The tool reads the scan and builds a new sheet. Your PDF stays as it was. Keep it: the scan is the proof, the sheet is the working copy.
Why does a long scan take minutes? Each page must be turned into a clear image and read word by word. A dense page with many rows takes real work. A one-page bill is fast; a thick pack of scans is a tea break. The sheet at the end is worth the wait — it is checked the same way as any text bill.
What about stamps and hand marks on the page? A “paid” stamp or a tick mark can cross the print and hide a digit. If a stamp sits on a sum, read that sum from the rows: add the rows and see if the sum you build matches the part of the stamp-free text. When in doubt, ask the vendor for a clean copy. Most will send one the same day.
What if a scan is in a second language? Print is print: sums, dates, and numbers read much the same. The tool reads the numbers and the layout. Check the words with more care, since names can blur in ways numbers do not. The totals check still guards the money.
Your next step
Find one scanned bill and run it through. Check the total and the bill number, then scan the rows. That is the whole habit — and it is the same first pass as how to check an invoice before you pay it, so a scan habit and a pay habit grow as one.
Then look at your pile with new eyes. The flat, sharp scans will read near clean. The bent photos are the ones to re-shoot. Sort those two stacks once, and scan work stops being a chore.
One more thought: ask your top vendors, once, to send text PDFs from the start. Each one who says yes takes a whole class of bills out of the scan path for good. The ask costs a line in an email. The gain runs each month, for years.
Sources and further reading
- Recognize text in scanned PDFs with Acrobat — Adobe explains how OCR turns a scan's picture of text into real text you can search and copy.
- Fix text recognition errors in scanned PDFs — Adobe's way to review the words a scan read with doubt and fix them — the same habit as checking a scan's rows.