Automated Accounts Receivable Processing Solution Proof of Concept (page 2)

TESTING STEP 2: Exception Handling

One method the OPEN SCAN solution uses to extract remittance data from document images is OCR (optical character recognition). Because OCR Software engines can make mistakes, the OPEN SCAN system has built-in validation processes designed to identify and correct errors. It compares the ‘amount paid’ portion of a remittance to the amount of the accompanying check to ensure that they balance to each other. In addition, the system utilizes a continually-updated validation file (which contains data pertaining to all open invoices) to verify the accuracy of data read from a payment document.

Once data has been extracted from the document, it is intelligently compared to the validation file using context-based algorithms. If the system determines with high confidence that the data is accurate, it is put into a file that will be loaded into the A/R system.

If the system cannot confidently validate the accuracy of the data, the entry in question is presented to an operator for verification or further research. At this point, the operator is presented with the image of the source documents (remittance and check). Also available to the operator is a search tool that allows them to view the validation file for comparison purposes.

In the company’s current work environment, they pay for 23,000 keystrokes and spend 13 FTE work hours to process and post 1,947 invoices. After automatically reading, validating and posting 1,536 invoices with no knowledge worker intervention. This left 259 (14%) lines of detail that required operator intervention to verify. The most common reasons for an invoice requiring assistance were:

1. Poor image quality
2. Item not found in validation file*
3. Insufficient information on the remit
4. Bad OCR read
5. Customer intention unclear
   *Test covered only one division, so a more complete validation file will address this issue.

There were 116 additional cases where the total remittance amount did not balance to the check amount due to an OCR mis-read in the amount column of the remit. These were tracked separately from the low-confidence posting items. In many cases the amount of OCR problems were concurrent with invoice matching issues. In all, the operator was able to view and verify all 259 questionable invoices, correct all 116 amount OCR errors, and manually enter all 158 handwritten items in 260 minutes.

Result 533 exceptions / 260 minutes = 67% reduction in time to process

In addition to reducing the hours spent by in-house staff on exceptions, 100% of bank lockbox keystroke fees were eliminated.

VALIDATION ACCURACY

At the end of the test, an output file including A/R posting data, like invoice number and net amount paid, was generated by the system. The output file from the system was compared to the output file that was originally created by the bank lockbox. The original data on the bank archival CD reported 1,947 lines of detail. The OPEN SCAN output file included 1,956 lines of
detail. The difference in these two numbers comes from the fact that operators using the system have access to the validation file and could quickly include details that the customer intended, but did not make clear on their documentation. Prior to the test, it was anticipated that a successful test may indeed generate more detail than the bank file. The ability to extract more information at the beginning of the process is expected to reduce the overall cost of the operation.

TEST RESULT COMPARISON (Metrics)

Page 1

Contact OPEN SCAN® today and find out how we can help you reduce costs while improving the efficiency of your receivables process.