First, confirm your PDF is actually scanned
Open the PDF and try to select a line of text with your cursor. If the text highlights and you can copy it, the file is digital (text-based) and you don't need OCR at all — you can extract the data directly. If nothing highlights and you're just drawing a box over an image, it's scanned.
Another tell: scanned pages often look slightly skewed, have speckles or shadows, or show different font weights where the paper was photographed or faxed. Statements downloaded straight from online banking are usually digital; anything that passed through a scanner, phone camera, or fax machine is usually an image.
This matters because OCR introduces errors that direct text extraction does not. If your file is already digital, skip OCR entirely and use a direct converter.
Run OCR to recover the text
OCR software analyzes the image and outputs recognized text, sometimes preserving rough layout. Common options include Adobe Acrobat's built-in Scan and OCR tool, free engines like Tesseract, and various online converters. Whichever you use, the goal is the same: produce a version of the file where the numbers exist as real characters.
Quality of input decides quality of output. A crisp 300 DPI scan, straight and well-lit, reads far better than a blurry phone photo. If you can rescan, do it flat and in good light. If you can't, deskew and increase contrast before running OCR — most tools offer a cleanup or enhance step.
Tables are the hard part. OCR reads text well but often struggles to keep columns aligned, especially with tight spacing or thin gridlines. Expect the amount column to sometimes merge with the description, or a date to split across cells. That's normal and fixable in the next step.
Move the result into Excel and rebuild the columns
If your OCR tool can export directly to Excel or CSV, start there. Open the file and check that each transaction sits on its own row with date, description, and amount in separate columns. If the tool only gives you plain text, paste it into a column in Excel, then use Data, Text to Columns to split on spaces or a delimiter.
Watch for the classic OCR misreads: a lowercase l or capital I read as 1, O read as 0, S read as 5, B read as 8, and commas swapped with periods in amounts. Decimal errors are the dangerous ones because they change values by orders of magnitude, so scan the amount column specifically.
Set your amount column to a number or currency format and confirm negatives came through correctly — parentheses or trailing minus signs are sometimes dropped. Add a header row and remove any repeated page headers, footers, or running-balance lines that snuck in as transactions.
Verify the numbers before you trust them
Never post OCR output to your accounting system without a reconciliation check. The fastest test: sum the debits and credits in Excel and compare against the totals printed on the statement. If they match, your extraction is very likely clean. If they don't, the difference often points you straight to the misread row.
Also check the opening and closing balances. Take the opening balance, add credits, subtract debits, and you should land on the closing balance. A mismatch means a missing, duplicated, or misread transaction.
Spot-check a handful of rows against the original PDF, paying attention to any amount that looks unusually round or unusually large. Ten minutes of checking here saves hours of chasing a reconciliation error later.
If you convert statements regularly, a purpose-built statement converter can save the OCR-and-cleanup steps by detecting columns and producing Excel, CSV, or accounting formats like QBO, OFX, and QIF directly. Whatever route you take, keep the balance-check habit.