Avoiding or reducing non-value-added activities in an organization can itself contribute to improved productivity. Such activities are present in many business processes, and organizations strive to address these using various measures like the process study, technology, lean methodology etc. Identifying and discarding duplicate invoices is one such activity in any accounts payable (AP) process. It has been observed that in a typical organization with more than 100,000 invoices per annum, there is, on an average, duplicate about 4-6% of overall invoice inflow is discarded every month. This not only increase the overall cost of processing invoices but also leads to potential duplicate payments.
Organizations receive accounts payable invoices from suppliers via multiple sources like emails, PO boxes, EDI/XML files, mobile apps, government portals etc. for processing and subsequently, ensuring on-time payment to suppliers. Suppliers often send the same invoice(s) via multiple channels to ensure that the invoice is delivered and processed quickly by their customers. This is the primary reason that creates duplicate invoices.
Duplicate invoices are typically identified and actioned-based on standard system checks using invoice parameters like business unit/company code, invoice date, invoice number, invoice amount, invoice currency, supplier number and supplier site. These parameters may vary depending on the organization’s policy or geographical requirements. However, this method has an inherent flaw; due to OCR inefficiencies or mistakes in manual data entries of these invoices, there are always chances that the data value(s) populated may have incorrect value(s). In such cases, the standard system check will not identify the invoice as a duplicate invoice, and it may get posted for payment. This also presents a considerable risk of fraudulent transactions.
The Traditional Approach
To overcome this issue, post-invoice processing, organizations perform an offline check at the payment stage, like manual checks, excel macros, or using a third-party analytics tool, to identify probable duplicates that are not caught during system check. This approach will work for organizations with limited, say weekly or bi-weekly payment runs. On the other side, organizations not deploying such methods end up putting additional effort in recovering the incorrect payments issued to suppliers. The monthly supplier AP statement reconciliation activity (wherein duplicate payments are identified, reconciled and payments are tallied, and for excess payments, debit notes are raised on suppliers for recovery) will result in open balances and extensive coordination with suppliers for the recoveries.
The Way Forward
Consolidating multiple inflow channels into one or two will result in reducing duplicate invoices. However, this is a major change management activity, and communicating and convincing your suppliers will be challenging. A simple solution for identifying near-match or probable duplicates at the invoice processing stage itself is desired by many AP organizations. There is a new breed of applications that use artificial intelligence and machine learning capabilities to predict near-match duplicates. These applications provide a warning or send an alert message to AP processors to take the necessary corrective action. In addition to traditional parameters like business unit/company code, invoice date, invoice number, invoice amount, invoice currency, supplier number and supplier site, other parameters like payment terms, currency, invoice line details such as UOM, unit price, item code etc., are included for more accurate prediction. The application may predict multiple duplicates for any given invoice based anything on the above parameters. To ensure only optimum effort by the AP processor to validate these alerts, often, a minimum confidence level threshold is set up - like for example, a prediction of anything above 90% confidence level should only be displayed to users. This threshold can be increased to higher levels based on the maturity achieved. There is an in-built, self-learning mechanism in the application to identify trends for specific invoice parameters and make more and more accurate predictions with time.
Using the AI-enabled, near-match duplicate invoice prediction feature, the AP team can focus on core activities and ensure the timely processing of payments to suppliers. This will save hours of effort in fixing (un-match/un-post) duplicate invoices and supplier follow-ups for recoveries, which in turn may save dollars spent on efforts and prevent incorrect payments. As per an in-house time study exercise conducted on process activities, on an average, unmatching and un-posting requires 15 minutes for each invoice, which essentially means avoiding 1,000 such unnecessary transactions will result in saving 250-hour worth of efforts. Most technology-driven organizations are adopting this approach to eliminate making duplicate payments. This, in turn, can also complement their green initiatives; for instance, a computer/laptop uses between 50-100 watts electricity in eight hours, and avoiding 1,000 duplicate invoices will save around 1,500-3,000 watts electricity. This can be a win-win solution for both customer and supplier organizations.