Optimizing your Accounts Payable process for OCR goes far beyond just asking your vendors to redesign their invoices for OCR. Here are five key steps to optimize your AP process for OCR:
- Capture the documents as early as possible in the process and in their original format
One of our clients was receiving invoices from vendors via e-mail. Because they did not have a solution in place to automatically capture those emails, they were printing out the invoice, only to turn around and scan the paper. I’m sure you can imagine the wasted time and money (and the environmental impact) of printing and scanning content that already existed in high-quality digital form.
- Work with your vendors to submit high quality documents
Have you ever heard of the term “garbage in, garbage out”? Yes, that may sound harsh, but the reality is that the accuracy of any OCR application will be highly dependent on the quality of images being fed through it. If you are submitting low-quality images to begin with, you should realize that your OCR results won’t be good. What is the right answer? Ask your vendors to submit images that are 300 DPI resolution (best practice for OCR). Most companies generate invoices electronically anyway, so ask them to send you the electronic copy generated by their invoice system (if they print and scan the invoice, which some companies are still doing, that will lead to an immediate loss in quality).
- Redesign invoices
This may be a harder sell, but you may want to ask your highest-volume vendors to redesign their invoices if you see any problematic areas. Generally speaking, you want to make sure that the invoices do not have shading on the document (especially in areas where you need to extract data). If possible, you will want to have the Invoice Total, Invoice Number, PO Number, etc. in the same location every time – and preferably all on the first page (if you are receiving multi-page invoices).
- Validate OCR data against your system of record to improve confidence
To get the best results, you need a way to validate the data that is being captured via OCR. For example, if you are trying to OCR a purchase order number, you should try to validate the purchase order number against your purchase order database to make sure that it is valid and matches the vendor submitting the invoice. If it doesn’t match, you can flag it for review by a user before the invoice gets sent to your AP system.
- Have realistic expectations going into any OCR automation project.
This probably should be listed as the first step. If your users are expecting near 100% accuracy from any OCR initiative, the project will not succeed. Make sure that you educate your business stakeholders early and often about the expected success rates and the types of OCR errors that will need to be corrected during verification.
