Invoice matching is a critical prerequisite in the B2B payment process but it’s hampered by slow manual processes. Many accounting professionals have a strong desire to improve the process — as many as 43.8% would like automated matching implemented.
In this article, you’ll understand the basics of invoice matching, different ways of implementing it, and its importance to a business. You’ll learn about the challenges of doing it manually and the benefits of fully automating it. Finally, you’ll see how fully automated invoice matching works under the hood.
Matching invoices is an important accounting task that compares the information in four types of documents to verify that they are all consistent with each other. The documents are:
When the accounting department receives a supplier invoice, the accounts payable (AP) team is responsible for matching invoice details against the details in the purchase order.
Some companies also match both against supporting documents like the goods receipt and inspection report, especially if the goods have high value or are supposed to meet stringent quality criteria.
If there’s a mismatch in the details of any two documents, it’s called a “match exception” or “deviation.” Mismatches can be of two types:
In the real world, there may always be tiny discrepancies in price or quantity. Prices may deviate due to taxes or currency exchange rates. Quantities may deviate due to unexpected damage or transport mishaps. In a shipment of 1,000 items, if 995 are fine and five are damaged, it may not be a wise choice for a supply chain to reject all 1,000 items.
To cater to these realities and avoid payment delays, businesses define some tolerances. If the deviations are within those tolerances, the documents being compared are treated as a successful match, and payment is approved. If the deviations are above the tolerance levels, the invoice is put on hold and sent back to the vendor for replacement, clarification, or rectification.
Let’s explore the different ways these documents are matched.
Depending on your business practices, industry, and local laws, there are four ways to select and match these documents with each other.
In two-way matching, the AP department verifies the vendor’s invoice against the corresponding purchase order. The items, prices, and quantities in the invoice should match those in the purchase order within acceptable tolerances.
Such PO matching is preferred when receipts and inspections are impractical or cost more than the goods themselves. It’s suitable for:
In three-way matching, accounts payable team compares the details in three separate documents:
This is the most popular method in most industries.
Four-way matching is like three-way matching but includes a fourth document in the comparison: the inspection report by the accepting department. The accepting department inspects the goods for quality and conformity to requirements. Only accepted goods are cleared for payment.
This method is ideal for industries like automotive and manufacturing where quality control is critical.
Some transactions, especially involving services, use contracts instead of purchase orders. The contract specifies the services, deliverables, metrics, and payment terms that have been agreed upon between the parties.
Because the information in a contract tends to be domain-specific and unstructured, it can’t be matched with an invoice mechanically. Instead, AP typically consults with the receiving department to understand the extent to which each term has been fulfilled and checks if the vendor has billed them correctly.
If the vendor has charged for a service that hasn’t been provided or fulfilled satisfactorily, the invoice is put on hold and sent back to the vendor for clarification.
Invoice matching, when done accurately, benefits many of your business practices.
First, invoice matching is recommended by accounting standards like the Generally Accepted Accounting Practices and the International Financial Reporting Standards. Your business will be compliant with all financial auditing and reporting requirements.
Second, invoice matching helps you keep an eye on your bottom line and avoid financial losses from invoicing mistakes or fraud.
And finally, invoice matching reduces the risks of legal actions or penalties from vendors.
Invoice matching remains a largely manual process because until recently, automated approaches have not been very reliable or accurate. Older automated approaches have problems understanding the wide variety of layouts and information in accounting documents like invoices and purchase orders. Companies are forced to rely on manual matching.
Unfortunately, doing it manually has challenges that cancel out many of the benefits of invoice matching:
Fully automated invoice matching combines deep learning, computer vision, and natural language processing to understand invoices and other documents the same way people do regardless of variations in layouts, terminologies, languages, or other aspects.
Older semi-automated workflows involve some manual steps like creating invoice templates for data extraction, entering data in spreadsheets before the automated steps take over, or verifying the extracted data after the automated steps.
Unlike these workflows, fully automated invoice matching is done by machines with minimal manual intervention. Let’s understand all the benefits of such AP automation.
Fully automated invoice matching replicates the deep semantic understanding of documents that people have. This allows accurate semantic matching of information across documents if there are variations in headings, words, terminologies, languages, currency formats, or date formats.
For example, if there are special conditions in both the purchase order and invoice, it can interpret those conditions just like people can and check if the items in an invoice match those conditions.
Even 4-way matching can be done in a matter of seconds compared to the hours (or even days!) that manual matching requires. The entire process is highly scalable because it’s automated end-to-end and supports millions of invoices per day.
Fully automated invoice matching lowers both your capital and operating expenditures. The capital expenditure on the invoice matching software and infrastructure is a lot lower than that required for hiring and training people. The operating expenditure on cloud infrastructure is far lower than ongoing employment costs. The cost per invoice also reduces by about 75%.
The accuracies of both the data extraction process and the semantic matching of information across documents are extremely high compared to both manual and older semi-automated approaches.
Unlike people, fully automated invoice matching can sift through all your historical invoices in seconds. This enables it to detect overpayments, duplicate payments, and other kinds of vendor fraud that are spread out over time.
Aside from vendor fraud, some medium and large businesses face insider fraud where employees collude with vendors for mutual benefits. Dishonest employees may cooperate with fraudulent vendors to tamper with purchase orders, hide overpayments, or allow duplicate payments. Fully automated invoice matching helps detect them.
The reliability and accuracy of the matching bestow a high degree of confidence in its results. Your financial audits are guaranteed to go far more smoothly with such accurate results.
The ability of fully automated invoice matching to complete even 4-way matching within seconds enables your business intelligence teams to get accurate cash flow and expense data in real-time.
A fast, accurate, and deep understanding of invoices enables you to avoid payment delays and fulfill all payment criteria of vendors. You may even receive early payment discounts from vendors. Accounts payable invoice matching strengthens your relationships with your suppliers.
Let’s understand some of the state-of-the-art techniques under the hood of fully automated invoice matching.
The first step involves accurate, deep, and semantic data extraction from all the relevant documents like invoices, purchase orders, goods received notes, and inspection reports.
This is achieved by combining deep learning, large language models, and text recognition models to identify and understand both printed and handwritten text just like people do.
Regardless of layout or other aspects, all the data in these documents are converted to structured data consisting of fields and their values. For example:
Fully automated invoice matching is capable of identifying tables, columns, rows, and cells across multi-page documents. Each invoice line item in a table is extracted and labeled based on its respective column.
Matching the products or services across invoices, purchase orders, and other documents can get complicated because:
Fully automated invoice matching uses intelligent data matching based on deep learning and natural language processing to solve all these challenges. Product titles, descriptions, and data like SKU numbers are matched intelligently using state-of-the-art deep large language models like GPT-3 to achieve human-level understanding.
Quantities and prices are matched regardless of currency and number localization. The tolerance levels for matching can be checked through price outlier detection techniques.
Payment and contract terms are typically written in natural language sentences that don’t conform to any strict syntax. While they can be very challenging for semi-automated approaches, fully automated invoice matching extracts and understands them effortlessly using large language models.
In this article, you saw how invoice matching benefits your business but how manual matching cancels some of those benefits. By using Width.ai’s fully automated invoice processing solution, you can process millions of invoices, matching them within seconds. Contact us to see a demo of our touchless invoice processing.
