One of the key aspects for the automatic invoicing process with Vendor Invoice Management (VIM) is to accurately extract information using OCR from invoice images. Without a high percentage of invoices being correctly extracted, we won’t achieve a high percentage of invoices being automatically processed.
Tools to improve OCR efficiency
To enhance the efficiency of extracting data from supplier invoices, carried out by OCR solutions like Opentext’s Intelligent Capture or Business Center Capture, there are various tools available that allow us to:
- Enhance information stored in the vendor master.
- Modify information in staging tables.
- Train invoices.
- Define new client rules for information extraction.
- Adjust the quality of invoice scanning.
- All the above tools are ineffective if we don’t know where to apply them.
How can we do this?
Before initiating any improvement process, it’s crucial to accurately identify what’s failing to decide where to start. A report providing real data on errors is necessary, rather than relying solely on user experience or perception of the most frequent errors. This report also enables us to measure the effectiveness of implemented improvements.
Since information extraction relies on the format of invoices generated by suppliers, the report needs to generate KPIs by supplier, such as:
- % of invoices with errors
- % of erroneous fields
- % of errors per field
With this information, we can quantitatively measure extraction failures and prioritize the most efficient improvements.
The standard VIM solution lacks a report showing aggregated information by supplier. In the Information Extraction Service, configuring IES Analytics is possible, but it doesn’t include aggregated supplier information.
OCR efficiency report
At ImagineRight, we’ve developed a report that measures OCR efficiency and generates KPIs from the information stored by the standard. The report is independent of the adopted extraction solution and works for both Business Center Capture (BCC) and Information Extraction Service (IES).
The report consists of two levels of aggregation:
- Aggregation by supplier
- Detail by invoice
In the supplier aggregation, defined KPIs are displayed, along with the total number of invoices and fields extracted.
At a detailed level, we can compare the value extracted by OCR with the correctly validated value by identifying invoices with extraction errors.
The report allows navigation to the VIM document to view the document image.
Improvement process
There are numerous factors influencing data extraction using OCR, hence, improvements to implement can’t be predefined. Prior analysis is necessary to identify areas for improvement. The steps to follow for implementing OCR data extraction improvements are:
- Define a starting point by selecting an initial period.
- Execute the OCR Efficiency report developed by ImagineRight.
- Identify fields with the highest number of extraction errors.
- For these fields, select the suppliers responsible for 80% of the errors.
- Implement improvements best suited to the selected suppliers.
- Evaluate the efficiency report after a period with implemented improvements.
- If the solution doesn’t meet expectations, iterate the process again.
As mentioned earlier, a high degree of autoposting depends not only on a high percentage of correctly extracted invoices. Following improvements in OCR, we are now able to address an automated invoicing improvement process in VIM.
At ImagineRight, we assist in optimizing the efficiency of VIM Opentext OCR on SAP. Our personalized OCR efficiency report provides a detailed insight into extraction errors, facilitating identification of improvement areas and prioritization of actions to achieve the best outcomes. Contact us for further information.
