What is OCR (scan and recognise)?
In short
OCR (optical character recognition) is text recognition: software converts a scan, photo or PDF into text a computer can read. With scan and recognise, the software then extracts fields such as supplier, invoice number, date, amount and VAT from an invoice, so you do not have to retype them.
How does scan and recognise work?
A purchase invoice arrives as a PDF, photo or on paper. The software first converts the image into text (OCR) and looks for the fields you need: the supplier, the invoice number, the date, the amount and the VAT. It then proposes an entry, with the general ledger account you used for that supplier last time. You or a colleague checks and approves it.
Software differs in what it recognises:
- Header data: supplier, date, number, total amount and VAT. That is enough for most invoices for services.
- Invoice lines: each item with quantity and price, needed if you want to match invoices against purchase orders and receipts.
More and more packages use AI models alongside classic OCR, so unusual layouts are read better too. A recognition remains a proposal: checking is still needed.
OCR or an e-invoice?
OCR is a detour: the supplier already had the data in its system, turned it into a PDF, and you extract it again. With an e-invoice in UBL, for example via Peppol, the data arrives without errors. So ask suppliers who can to send e-invoices, and use scan and recognise for the rest.
What should you look out for?
Test with your own invoices, including a few tricky ones: several VAT rates, a foreign supplier or a poor scan. Read more in automated invoice processing: how to choose software.
Which software helps with this?
Many accounting packages recognise an invoice's header data themselves. Specialist scan and recognise software also recognises invoice lines and matches them to purchase orders.
Compare in the category
Including Zenvoices, 4CEE (voorheen TradeInterop) and Blue10.