A client hands you a pile of paper statements, or sends photos taken on a phone. There's nothing to copy and paste, because the PDF is really a picture of a page. Getting that into Excel takes a different approach from a normal digital statement, and it needs a check at the end, because scans are where reading errors creep in.
This guide covers how to tell whether you have a scan, how to get a good one, how to convert it, and how to prove the numbers came across correctly.
Working with a normal, digital PDF statement? Start with how to convert a bank statement PDF to Excel. For OCR on documents in general, see how to extract text from scanned PDFs.
Is your statement a scan or a real PDF?
Open the PDF and try to select a line of text with your cursor.
- You can select and copy words: it's a digital PDF. Use any method in our PDF to Excel guide.
- You can't select anything, or you select the whole page as one block: it's a scan. The "text" is a picture, so you need OCR (optical character recognition), software that reads the picture and turns it into text.
Photos from a phone camera, scanned paper statements and "printed to PDF" images are all in the second group.
Step 1: Get the best scan you can
The quality of the scan decides how reliable everything after it is. A few minutes here saves hours later.
- Lay the page flat. A curled corner or a fold bends the text and confuses the reader.
- Use even light with no shadows. Hold your phone directly above the page, not at an angle. Avoid glare from a window or lamp.
- Scan at 300 dpi or higher if you're using a scanner. Lower settings blur small print, and bank statements use small print.
- Capture the whole page, edges included. Cut-off numbers at the margin are the commonest cause of wrong totals.
- Keep pages in order and keep them together. One PDF per statement, with the pages in sequence.
- Don't use filters or "enhance" modes that sharpen or darken the page. They can turn a 3 into an 8.
Most phone scanning apps (the built-in scanner in your phone's notes or files app, for example) can produce a flattened, cropped PDF.
Step 2: Convert it with OCR
You have three options:
- Excel's PDF import works only on digital PDFs. It can't read a scan.
- A general OCR tool can turn the page into text, but you then have to rebuild the table yourself: which numbers are dates, which are money in or out, which line belongs to which transaction.
- A converter built for bank statements does the OCR and rebuilds the transaction table in one step.
With PDFDrill, you upload the scanned statements, which can be a batch for one client, and it reads each page, extracts every transaction into a table, and checks that the statement balances.
Step 3: Check the numbers (do not skip this)
OCR is good, but it isn't perfect. On a clean scan it often gets everything right. On a faded, creased or low-resolution one, it can confuse similar-looking digits (3 and 8, 1 and 7, 0 and 6) or miss a line entirely. The mistake is almost invisible when you read through the spreadsheet, and it can matter a lot: a misread amount silently changes the client's books.
The simplest check is the balance check:
opening balance + all transactions = closing balance
Do it for every statement:
| Cell | What goes in it |
|---|---|
| F1 | The opening balance printed on the statement |
| F2 | The closing balance printed on the statement |
| F3 | =F1+SUM(C:C) (money in positive, money out negative, in column C) |
| F4 | =F3-F2 |
If F4 is 0.00, the amounts reconcile, but that does not prove every row is correct: two errors can offset each other. Compare the extracted rows with the source as well. If it isn't zero, the difference tells you where to look:
- A difference equal to one transaction's amount: that row was missed, or counted twice.
- A difference of 5.00, 50.00 or 0.50: a single digit was probably misread in one amount. Look for one that has an 8, a 3, a 6 or a 5 in it.
- A difference that's a round number in a different place: a decimal point may have been dropped or added.
PDFDrill runs this check automatically on each statement and shows PASS or FAIL with the exact difference. It also checks that one month's closing balance matches the next month's opening balance, so you spot a missing page or statement. Rows it flags can be viewed alongside the source page, edited, and the checks re-run.
What the balance check can't catch: a misread description (a payee spelled wrongly) doesn't change any total. Skim the descriptions of larger transactions, and check dates, which can also be misread.
Common problems with scanned statements
Numbers ending up in the wrong column (money in vs money out). Statements with separate Debit and Credit columns are the commonest problem on scans. The balance check catches it, because swapping a column changes the total.
A blank or missing page. The statement won't balance. Compare page numbers ("Page 2 of 4") and the first and last balance on each page.
Statements that overlap. If two scans cover some of the same days, the same transactions appear twice. PDFDrill marks these as duplicates, so you can see exactly which rows match.
A photo taken at an angle. Retake it. Straightening software helps, but a flat scan is much better.
Handwriting or stamps on the page. They can be read as stray characters in the table. Remove any rows that aren't transactions.
FAQ
Can OCR read a scanned bank statement accurately? OCR works well on clean, flat, high-resolution scans and worse on faded, skewed or low-quality ones. Because it can make mistakes, always run the balance check: a misread amount can show up as a difference between the calculated and printed closing balance. Also compare the rows with the source, since offsetting errors can leave the total unchanged.
What resolution should I scan at? 300 dpi is a good minimum for statements. Higher gives more detail, at the cost of bigger files.
Can I convert a phone photo of a statement? Yes, if it's flat, well lit, in focus and shows the full page. A scanning app that flattens and crops the page usually gives better results than a plain photo.
Is it safe to upload scanned statements? It's a fair question to ask of any tool. With PDFDrill, files are stored privately. Source PDFs are deleted 7 days after the batch is created and extracted data after 30 days, and you can delete a batch at any time. Details are on the security page.
Can I get the data in CSV for Xero or QuickBooks? Yes. See how to convert a bank statement PDF to CSV and how to import a bank statement into Xero.
Got a pile of scanned statements? Try PDFDrill on one: it reads scans, extracts the transactions and checks every statement balances. It's free for your first 10 pages each month.