Recognition Accuracy Monitoring and Improvement

During testing, review recognition accuracy for high-volume suppliers, frequently used invoice layouts, and invoices that require repeated manual correction.

Focus on attributes that directly affect invoice creation, routing, matching, and exception handling.

Review area	What to monitor	How to improve
Supplier recognition	Supplier name, supplier site, supplier address, and duplicate or ambiguous supplier records.	Review supplier master data, correct supplier/site details, and remove avoidable ambiguity where possible.
Purchase order recognition	Purchase order number recognition, PO format consistency, and PO matching behavior.	Review PO number formats and supplier invoice presentation. Alphanumeric and distinctive PO formats generally improve recognition.
Business Unit derivation	Business Unit assignment for matched invoices, unmatched invoices, sender email mapping, recipient email mapping, and routing attributes.	Correct Business Unit derivation setup, email mappings, or routing attribute usage.
Header attributes	Invoice number, invoice date, invoice amount, currency, payment terms, legal entity, and requester.	Correct recognition errors and save learning where supported. Review requester domain setup if requester recognition fails.
Line attributes	Description, quantity, unit price, line amount, purchase order number, and ship-to location.	Review line recognition options, service order settings, utility supplier settings, and adaptive learning results.
Invoice image quality	Blurred scans, skewed images, incorrect orientation, low contrast, cropped pages, or unsupported files.	Correct scanner settings, enforce file submission standards, and retest with cleaner invoice samples.
Training coverage	Suppliers or layouts with repeated recognition issues.	Use Invoice Document Training for high-volume suppliers, recurring layouts, and suppliers with high exception volumes.
Adaptive learning	Whether corrections are reused for future invoices with the same layout.	Confirm corrections are saved, fingerprints are generated, and later invoices show improved recognition.
Bulk reprocessing	Whether applicable incomplete invoices with the same fingerprint receive the saved correction.	Verify the invoices share the same fingerprint and that the correction is applicable. Review any invoices that remain unchanged individually.
Reporting trends	Suppliers with low recognition rates, recurring exception patterns, and attributes with frequent corrections.	Use recognition reports to prioritize supplier training, supplier data cleanup, or configuration changes.

Use the Invoice Documents Recognition Tracking Report to review invoice-processing progress and import outcomes. Use the Invoice Documents Recognition Rate Report to identify suppliers, layouts, and attributes that need additional review or training. After corrections, training, or configuration updates are completed, process additional invoices from the same supplier or layout and confirm that recognition accuracy improves.