Extracting a scanned Maybank statement to Excel
Plenty of Maybank statements reach an accounts desk as scans: a printout from the branch, stamped and signed, run through the office copier and emailed as a PDF. Parserje reads those into Excel. How it does that is worth five minutes, because it explains both why a scan takes longer and why the soft copy, when you can still get it, is the better file.
Two kinds of PDF
A Maybank statement downloaded from Maybank2u carries a text layer. Every date, description and amount is stored in the file as text, and Parserje reads that text together with where each word sits on the page. The figures come across exactly as the bank printed them.
A scanned statement has no text at all. Each page is a picture of paper. To get figures out of it, the page has to be read by optical character recognition, or OCR: software that looks at the shapes on the page and decides which characters they are.
You do not have to tell Parserje which one you have. It checks every page for readable text, and only the pages without any are sent to OCR. A file that mixes the two is fine.
What happens to a scanned page
The OCR runs in your browser, on your own computer, like the rest of the conversion. Parserje reads each page, finds the transaction table on it, and pulls out the date, description, amount and balance for every row.
Then it checks what it read. Each row is tested against the running balance: the balance on the line before, plus or minus this amount, has to equal the balance printed on this line. Each statement is tested against the beginning balance, ending balance and totals printed on it. When an amount fails that test and the balance column shows what it must have been, the amount is corrected. When Parserje cannot be sure, it flags the row instead of guessing.
What is particular to Maybank scans
- The branch stamp. A certified copy often carries an officer's stamp and signature across the table. Parserje keeps the stamp's words out of the transaction descriptions.
- Amounts under one ringgit. Maybank prints them without a leading zero, so the decimal point is the first thing on the line. On a scan that lone dot is the easiest mark on the page to lose, which is why these are the rows most often flagged.
- Rows the OCR drops. A faint line can vanish from the OCR output entirely. Because every balance has to follow from the one before it, a missing row shows up as a balance that does not add up, and Parserje uses that gap to recover the row or to flag the spot where it went.
Why the soft copy is faster and more accurate
Faster. Reading a text layer takes a fraction of a second a page. OCR takes a few seconds a page, so a ten page scan takes about a minute and a year of scanned statements takes several. Parserje tells you how long a scan will take and asks before it starts a long one.
More accurate. The figures in a text layer are the bank's own, so there is nothing to misrecognise. A scan adds a chance of error at every step between the bank and you: faint print, a crooked page, copier noise, a stamp across the figures. Any of them can turn an 8 into a 3 or swallow a decimal point. The checks catch most of that, but a check that never has to fire is better than one that does.
So if the month is still available in Maybank2u, download it there and convert that instead of scanning the printout. Keep the scanner for the months where the soft copy is gone, or where the client only kept paper.
Getting the best from a scan
- Scan at 200 dpi or better, flat on the glass, straight, in one pass. A phone photo taken at an angle is the hardest case there is.
- Scan every page in full, including the bottom of the last one. The ending balance and the totals printed there are what the statement is checked against.
- Keep one account per file, with the pages in order.
After it reads
Compare the closing balance in the result with the ending balance printed on the statement, then look at any flagged rows against the PDF. Those two habits, and four more, are in getting an accurate conversion every time. New to Parserje? Start with how to convert a bank statement PDF to Excel.
Questions people ask
Can Parserje read a scanned Maybank statement?
Yes. When a page has no text layer, Parserje reads it as an image with OCR, inside your browser. A clean, straight scan at 200 dpi or better gives the best result.
How long does a scanned statement take?
A few seconds a page on a typical laptop, so a ten page scan takes about a minute. The soft copy of the same statement takes seconds. Parserje tells you before a long scan starts.
How do I know if my PDF is a scan or a soft copy?
Open it and try to select a line of text, or search for an amount with Ctrl+F. If the search finds it, the PDF has a text layer and will read quickly. If nothing can be selected, it is a scan.
Convert your bank statement to Excel
Drop in your PDFs. Parserje reads them, checks the figures against the statement, and tells you what it found.