In this article:
Matching Overview
As payments are created via Cash Application, Tesorio generates match recommendations for both customers and invoices. Where match confidence is high, drafted payments are automatically applied to customers and invoices. Tesorio uses the following information to identify matches
Memo field on the payment transaction record
Amount of the payment
Customer name
Customer aliases
Invoice amount
Invoice numbers
In making payment recommendations, Tesorio runs all of the rules outlined below for every transaction. Then, for a given payment, we combine all recommendations that belong to the same customer or invoice, and assign a final score. In this way, payments that match more than one rule (e.g. customer name + invoice amount) receive a higher confidence score and a higher ranking in our recommendation list. When this ranking is high enough, Tesorio applies the recommendation automatically to a payment.
Customer Identification
Tesorio generates customer match recommendations based on data found in the payment memo field. The data in the memo field is compared against the list of customer names from your ERP as well as any Aliases you have added on the customer record, and where commonalities are identified, a customer match is recommended. Often, an invoice match will accompany a customer name match.
Alias Names
In many cases, customer identifiers in the payments you receive are somewhat different than the customer name stored in your ERP. Add Aliases to your customer records in Tesorio to improve customer matching. Aliases can either be names (for example, a parent company paying on behalf of a child, or the legal entity name or DBA that differs from your customer name in the ERP) or they can be alphanumeric strings that serve as unique identifiers.
You can add aliases at any point from the Customer Details page,
Or during the Cash Application process, when matching to a customer, log aliases for future use.
Invoice Identification
To identify invoice matches, Tesorio reviews all open invoices as well as invoices that have been closed within the last 7 days. Though a payment cannot be applied to a payment that is closed or has a 0 balance, the algorithm still reviews these invoices because in some cases finding matches against these invoices highlights previous cash application errors.
Single Invoice Matches
Tesorio also generates invoice match recommendations by comparing the payment amount to the invoice balances in your account. When a match is generated based on the invoice amount, the panel also surfaces the invoices that triggered the recommendation so that the payment can be applied in full.
Payments that match open invoice amounts exactly will have the highest recommendation confidence. But, the algorithm will return matches for any invoice that is within a 10% tolerance of the payment amount in order to identify cases of underpayment and overpayment due to FX conversions, transaction fees, etc.
Multiple Invoice Matches
After reviewing individual invoices, Tesorio groups the invoices under consideration by customer and compares
The sum total of all of the customer's open balances against the payment amount to identify potential payment against multiple invoices. Similarly to the single invoice matches, a 10% tolerance is applied, so any sum of invoices within 10% of the payment amount will be returned as a recommendation, with exact matches ranked highest.
Combinations of the customer's open invoices, beginning with the oldest invoices, against the payment amount to identify potential payment against multiple invoices. Similarly to the single invoice matches, a 10% tolerance is applied, so any sum of invoices within 10% of the payment amount will be returned as a recommendation, with exact matches ranked highest.
Invoice Number Matches
Tesorio reviews all payment memos for references to invoice numbers. If an exact match for invoice number is identified, the customer and invoice are returned as an auto-match. In the payment memo, invoice numbers are often accompanied by prefixes that help Tesorio identify them. Common prefixes include:
Inv
Bill
No
Num
Number
If the customer recommendation is correct, but the invoice recommendation is not, users have the option to match to the customer only by clicking the customer link to the right of the name.
Tesorio Recommendations
Recommendations are generated automatically at the time that the bank feed is ingested with payment transaction data. If there are older unposted payments that have not yet been matched, recommendations are regenerated for them each time new transactions are imported to account for changes in the open invoices.
Tesorio will generate up to 10 recommendations for each payment. If no recommendation is available, customers can always manually select the customer and/or invoice the payment corresponds to by clicking the Unidentified customer name or selecting an invoice to apply in the Available section.
Note: Recommendations are not generated for manually created payments or for unapplied payments in the Posted Payments workspace.
Tesorio Auto-Match
Tesorio is constantly measuring the success rates of our recommendations to ensure quality. Each recommendation is assigned an internal score, and when the scores exceed our threshold, Tesorio automatically matches the payment to a customer and invoice. If there are multiple recommendations for the same payment that exceed the score threshold, the recommendation with the highest score is used for auto-match.
Regardless of whether payments received recommendations or auto-matches, users need to review all payments and move them to a 'Pending' status by clicking the Post button, individually or in bulk, in order to post the payments to the ERP.
AI and ML in Cash App Recommendations
Tesorio uses a combination of AI, machine learning, and rule-based algorithms to identify potential customer and invoice matches in Cash Application.
Customer Name Identification uses ML vectorization and similarity search to compare customer and alias names to data in the bank transaction memos. With this approach, the memo texts and customer names are broken into n-grams and then compared against customer names and aliases to identify the best possible matches.
Invoice Amount Matching is rule based. Following the criteria above, Tesorio identifies individual or combinations of invoices.
Invoice Number Identification uses three different algorithms.
The first uses heuristic based text matching to find common invoice patterns
The second uses an ML entity recognition and extraction model. This model identifies and extracts relevant invoice number information in the bank transaction memo. This model learns from the acceptance of previous matches, and compares the bank transaction memo lines to the invoice numbers of posted payments so that with each additional example of applied payments, the model improves.
The third model splits memos and AI extracted values by common word end markers (like spaces, hyphens, and special characters). It then removes all non-digit characters and looks to see how many changes would be required to go from the search string to the searched string (invoice data from your system).
Tesorio is constantly reviewing, updating, and improving our matching algorithms, so these approaches are subject to change as new approaches and use cases are evaluated.