You can convert a scanned PDF to Excel free with OCR, which is software that reads the picture of the page and turns it into text. Free OCR handles clean, straight scans of short documents. OCR can misread digits, which a text PDF never needs to guess, so check every total against the statement.
Why does a scanned PDF need OCR before Excel?
A scanned PDF needs OCR first because it contains a photograph of the page, not text. The file holds pixels. A converter cannot pull a date or an amount out of pixels until optical character recognition (OCR) has looked at the picture and decided which letters and digits it shows.
A text-based PDF is different. The bank's system wrote the characters into the file, so a converter reads them exactly, with no guessing. A scan adds a guessing step before any conversion starts, and every guess can be wrong. That is why the same statement can convert perfectly as a downloaded PDF and badly as a scan.
The chain for a scan has three links, and an error in any one carries forward:
- The scan itself. Resolution, skew, shadows, folds and fax-style noise all change what the OCR sees.
- The OCR step. The software turns shapes into characters and also guesses where each column begins and ends.
- The table step. Another layer groups the characters into rows and cells, then writes them into Excel.
Free tools often combine all three behind a single upload button, so you never see which link failed. You only see the result.
Is your statement PDF scanned or text-based?
You can tell in about ten seconds: open the PDF and try to select a line of text with your mouse. If the text highlights word by word, the file is text-based. If a block or the whole page highlights as a picture, or nothing highlights, the file is a scan.
Three quick checks give the same answer:
- Select and search. Press Ctrl+F (Cmd+F on a Mac) and search for a word you can see on the page, such as the bank's name. A text PDF finds it. A scan finds nothing.
- Zoom in hard. Zoom to 400 percent. Text stays sharp, since it is drawn from outlines. A scan turns blocky or blurry.
- Copy and paste. Paste a copied line into a text editor. A text PDF gives you the words. A scan gives you nothing, or random symbols.
A statement downloaded from your bank's website is almost always text-based. A statement that arrived by post and was photographed, or run through an office scanner, is a scan. Some scanners run OCR on their own and save a "searchable PDF". That file looks like a scan and has a hidden text layer, which may be wrong in places because someone else's OCR wrote it.
Why it matters before you choose a tool
If the file is text-based, you do not need OCR at all, and a plain converter will give better results. If you only have a scan, the cheaper fix is often to ask for the text version. The post on scanning paper statements well at 300 dpi covers the settings that help when a paper original is all you have.
CheapBankConvert is built for text-based PDF statements, according to its homepage FAQ, so a scan is not what it reads best. It is not yet open for conversions either, so this post stays with what free OCR does.
How far does free OCR go on a scanned PDF?
Free OCR goes as far as a short, clean, upright scan with a plain table. It tends to stop at a page cap, a file size limit or a daily quota, and it loses accuracy as the scan quality drops. Where exactly each tool stops differs, so read its limits before uploading.
Free OCR tools commonly limit some of these things:
- Pages per file or per day. A tool that allows a few pages cannot finish a ten-page statement.
- File size. Photographed pages are large, and a size cap may reject them.
- Languages and scripts. Free tools are strongest in the common ones, and weaker on unusual fonts.
- Layout. Tables with faint lines, colored rows or two amount columns are harder than a plain list.
- Output. Some give you Excel only after a signup, or return a file with a watermark.
What quality does it need?
Free OCR gives better results when the scan meets a few conditions. Treat them as a checklist before you upload:
- The page is straight, with no skew or curl at the edges.
- The scan is sharp enough that you can read small digits without squinting.
- The page is evenly lit, with no shadow across the amounts.
- The text is dark on a plain, light background.
- Nobody wrote over the table, and no stamp covers a number.
A phone photo of a page on a desk usually fails two or three of these. A flat scan at a proper resolution passes most. If free OCR returns a mess, the cause is often the scan, so rescan before you blame the tool.
The page cap trap
A common trap is a tool that converts the first few pages and then quietly stops, or asks for payment on page four. Count the rows in the result against the statement. If the sheet ends mid-month, the cap hit you. A slower and more honest tool tells you the limit before you upload; a tool that hides it wastes your time. For the broader pattern across free tools, see where free PDF to Excel tools stop.
Scanned statement OCR errors: 8s read as 3s, lost minus signs
OCR on a scanned statement makes a small set of repeat mistakes, and they cluster in the digits that matter most. A misread digit gives a wrong amount that still looks like a valid number, so nothing in the sheet signals it. You catch it only by checking against the statement.
The errors below are the common ones. They are examples of types, not measurements of any one tool.
| What the page shows | What OCR may return | Why it happens |
|---|---|---|
| 8 | 3 or 0 | Closed loops blur together on a soft scan |
| 5 | 6 or S | Curves look alike at low resolution |
| 1 | 7, l or I | A thin stroke, especially in a small font |
| Decimal point | Missing, or read as a comma | A small dot disappears in noise |
| Minus sign or "CR"/"DR" | Dropped | A short dash or faint letters vanish |
| Comma in 1,204.50 | Period, or a space | Similar marks at small size |
| 0 | O | Same shape in many fonts |
Which errors do the most damage?
Lost minus signs change the direction of the money. A withdrawal of 45.00 that loses its sign becomes a deposit, and the balance moves by 90.00 more than it should. The error is silent unless you check the totals.
Missing decimal points change the size. 1204.50 without its point becomes 120450, which a sum then reports as a huge number. This one is easy to spot because it breaks your totals loudly.
Swapped digits are the worst, because the new number looks believable. An 8 read as a 3 turns 1,208.00 into 1,203.00. Your sheet shows a plausible amount, and only the balance check will say otherwise.
Dates can be wrong too. A 5 read as a 6 can move a transaction by a month or put it outside the statement period.
Descriptions are less critical, but not free of errors
Payee names tolerate small mistakes when you read them, but a changed letter may stop a rule or a filter from matching. If you categorize by payee, scan the descriptions for odd characters such as a zero where a letter O belongs.
Checking free OCR output against the scanned statement balance
You prove OCR output right by tying it to the balance the bank printed. Opening balance plus the signed sum of the rows must equal the closing balance. If it does, almost any digit or sign error would have broken it, so the sheet is very likely correct.
Checklist for verifying OCR output
- Read the opening and closing balances from the page itself, not from the OCR result, and type them into two cells.
- Make sure every amount cell is a number: select the column and check that the status bar shows a Sum.
- Add the signed amounts, add the opening balance, and compare the result with the closing balance.
- If the statement prints a running balance, compare a few of its lines with the same line in the sheet, especially the first line of each page.
- Count rows page by page and compare each page's count with the sheet.
- Check signs: pick two withdrawals and two deposits on each page, and confirm each sits on the right side of zero.
- Check the first and last date against the statement period, and look for any date outside it.
- Look for a gap: a row count that is short on one page points to a dropped line.
A worked example
Take an example statement (the numbers are made up). The opening balance is 2,500.00 and the printed closing balance is 2,061.40. The OCR sheet has 21 rows, and their signed sum is -348.60, so the computed closing balance is 2,151.40.
The gap is 90.00 (2,151.40 minus 2,061.40). Because 90.00 is exactly twice 45.00, the likely cause is one withdrawal of 45.00 that lost its minus sign and was read as a deposit. You look for a 45.00 row on the statement, find it, and flip the sign. The balance now ties out.
If the gap is not a clean multiple of a row's amount, look for a swapped digit, a missed row, or a decimal error instead. The gap tells you what to look for, which is the real value of the check.
When the check fails and you cannot find the cause
Stop fixing and get a better input. Ask the bank for the text PDF, rescan at a higher quality, or move to a tool with fewer limits. If your next idea is a trial, read about skipping trials that bill you later first, so the search for a free OCR tool does not end in a subscription.