You can try to convert a handwritten PDF to Excel, but the result is only as good as the writing. Ordinary OCR is built for printed type and reads neat handwriting poorly. Plan on checking every figure, and on retyping the pages when the writing is untidy or the stack is small.
Can OCR read handwriting in a PDF?
Ordinary OCR reads printed type well and handwriting badly. It was designed around fonts, where every letter 'a' looks nearly the same. Handwriting varies from person to person, and from line to line within one page, so the software has far less to match against.
Separate tools exist for handwriting recognition, and some do better on neat writing. Their results depend heavily on the writer, the pen and the scan. Do not assume that a tool which handles a printed bank statement will cope with a handwritten ledger. Test it on one page before you commit the whole stack.
A few things hurt handwriting recognition the most:
- Joined-up writing, where letters run together and the software cannot tell where one ends.
- Slanted or uneven lines, which blur the rows.
- Pencil and light pen, which scan faintly.
- Crossed-out and overwritten entries, which leave two possible readings of one cell.
Also check whether your PDF is a picture at all. A handwritten page is always an image, so it needs a text layer before Excel can use it. The three-step check in how to tell if a PDF is an image applies here too, and how OCR builds columns from a scanned page describes the process that handwriting strains.
The converter does not warn you when it guesses. It will return a clean-looking cell with a wrong value. Some tools attach a confidence score to each reading, and what OCR confidence scores mean and when to distrust them explains how far to trust them. Neither a score nor a neat spreadsheet replaces a check against the page.
Handwritten numbers vs handwritten words
Handwritten numbers are usually easier to recognize than handwritten words. A digit has only ten possible answers, while a word has thousands, and a column of amounts follows a pattern the software can test. Even so, number errors cost more, because one wrong digit changes the money.
Words can be wrong and still be useful. A payee read as "Smth Plumbing" is easy to fix, and the meaning survives. A number read as 1,170 instead of 1,770 looks perfect and is wrong. For that reason, handwritten figures need more care than handwritten text, even though they are easier for the software.
Three habits make handwritten numbers safer:
- Add up the columns. If the page has a total, compare it with the sum in Excel. A mismatch means at least one cell is wrong.
- Check against a second source. A bank statement or invoice that covers the same period can confirm the key figures.
- Look for the classic confusions, such as 1 and 7, 4 and 9, and 5 and 6. The digit confusions that OCR makes on printed statements are the same ones that handwriting makes worse.
Boxed columns, like a ruled ledger with a column for each digit group, help a lot, because they separate the figures. Free-form lists, where amounts float beside descriptions, are the hardest.
Handwritten notes on a printed statement
Handwritten notes on a printed statement are usually ignored or misread, while the printed lines convert normally. The printed text is what the software is built to read, and a margin note is an unexpected mark. The risk is when the pen touches a printed figure.
Notes fall into three kinds:
- Margin notes, such as "paid by cheque 214" beside a line. These are generally outside the transaction columns and are skipped or return as stray text. Type them in yourself if you need them.
- Ticks and circles around amounts. They are harmless when they sit beside a figure, but a ring that crosses a digit can change how it reads.
- Corrections written over printed amounts. Treat these rows as unreadable. Check them by hand and take the written value as your source only if you trust it.
A statement with many notes over the transaction area is a poor input, and typing it is safer. For a page with only a few marks, convert the printed part and then add the notes to a new column by hand. A statement converter such as FastBankConvert is built for printed layouts, and its homepage makes no claim about handwriting, so keep handwritten pages out of that route.
When is retyping a handwritten ledger faster?
Retyping is faster when the page count is small, the writing is untidy, or the content is words rather than figures. Conversion only pays off when the writing is neat, the columns are boxed and the volume is large enough that checking is quicker than typing.
Use this checklist, and answer each question for your own pages:
- How many pages? A handful of pages is usually quicker to type than to convert and then verify. A bound ledger of hundreds of pages is where a tool might earn its time.
- How neat is the writing? Block capitals and clear digits convert far better than cursive.
- Numbers only, or words as well? Columns of figures convert more reliably than free text.
- Boxed columns, or free form? Ruled columns help. A free-form list does not.
- Can you verify the result? If the page has totals or a matching bank record, you can check a conversion. If not, you must read every cell, and that removes most of the time saved.
Count the checking in your estimate. Converting a page is fast, but verifying it means reading the page against the sheet line by line, and that time belongs in the comparison with typing. Two or three of the answers pointing to "convert" make a trial worthwhile, and any answer of "untidy" or "free form" points to retyping.
If the pages are printed and only scanned, the picture is different. Converting a scanned PDF to Excel covers printed scans, and turning a PDF back into an Excel file covers a sheet that was saved as PDF. FastBankConvert, built for printed bank statement PDFs, is getting ready to open, and the get started page takes your email.