SAP

Author

Samantha Wong

Samantha Wong

Samantha Wong is a Product Manager for Predictive Analytics at SAP. With a Bachelor of Commerce degree specialized in Marketing, and experience in Advertising, Banking, and Insurance, Samantha fits the profile of analyst turned citizen data scientist. For interesting articles and commentary on the ever-evolving world of Predictive Analytics and Machine Learning, connect with Samantha on Linkedin

Keep in touch

Subscribe for the latest news, updates, tips and more delivered right to your inbox.

Subscribe to updates

Category

Learning

Connect with us

Accounts Receivable (AR) is typically the largest asset to any organization's financial statements. With B2B transactions increasing in volume and complexity, poor management of AR can lead to unnecessary expenses and cash flow problems.

Using machine learning models built with SAP Analytics Cloud, the Payment Predictions Business Content learns from historical cleared invoices to predict when outstanding invoices will be paid. By identifying invoices that are likely to be paid late, Collections Managers can focus on customers who are likely to have large amounts of overdue receivables. This ultimately optimizes collection strategies, leading to lower Day Sales Outstanding (DSO) and better visibility into future cash flow.

Package Contents:

The SAP Finance: Accounts Receivable – Payment Forecasting business content contains 4 objects:

  • Story: SAP__FI_AR_IM_PAYMENTFORECASTING
  • A sample dataset of historical cleared invoices
  • A sample dataset of current invoices (outstanding)
  • A sample dataset containing the current invoices with predictions applied

This blog will explain how to navigate and understand the story’s contents. The sample datasets are included for users who want to recreate the predictive model and experiment. For more information on the steps to recreate this scenario, please see the documentation link at the bottom of the page.

Story Page: Invoice Overview

The first page in the story provides an overview of the current state of receivables as of November 28, 2018.

The key elements within this page are:

Overdue Receivables % and Outstanding Receivables
The Overdue Receivables % KPI is a common calculation which describes the amount of the invoices that are overdue as a percentage of the total open invoices. In this example, the total amount of overdue receivables ($33.377 million) divided by the total amount of open receivable ($85.807 million) is equal to 38.90%. Conditional formatting has been applied to this KPI where it appears: green if < 10%, yellow if >= 10% and < 25%, and red if >= 25%. This threshold can be configured in the story to reflect your business.

Overdue Receivables
The Top Overdue Customers chart shows the top 5 customers by their total amount of overdue receivables.

The Overdue by Days chart groups all overdue invoices into buckets which represent the number of days they are overdue. The total overdue amount is shown for each bucket. In this example, $7.011 million is overdue by 1 – 15 days. $12.85 million is overdue by 31 – 60 days.

Future Due in Receivables
Future Due in Receivables refers to the invoices that are open as of November 28, 2018, but are not yet due, meaning their due date is later than November 28, 2018.

The Incoming by Days chart groups all non-overdue invoices into buckets which represent the number of days until they become due. In this example, $27.857 million will become due in 0 – 15 days.

The Payment Predictions chart is based on predictions from a regression model built using Smart Predict in SAP Analytics Cloud. This regression model predicts the number of days it takes an invoice to be paid from the invoice date. From this prediction, we can derive whether the invoice will be paid on time or late, and by how many days because we can compare the prediction against due date information in S/4HANA. This chart shows the total amount of receivables that are predicted to be paid on time (the green bar) alongside the invoices predicted to be paid late (the red bars). The late invoices are grouped into buckets which represent the predicted number of late days. In this example, $26.141 million is predicted to be paid between 1 – 15 days late.

The two charts have linked analysis enabled. If we click on the bar representing receivables due in 0 – 15 days, the Payment Predictions chart will change to reflect this subset of data. In this example, of the $27.857 million due in, we predict $5.466 million will be paid on time, $14.432 million will be paid 1 – 15 days late, and so on.

Outstanding Receivables Detailed View

This table shows information about individual invoices and is controlled by the input control on the left titled Overdue (Days). The default view contains all invoices that are currently overdue (the number of overdue days is >0). To see non-overdue invoices, change the input control to only select invoices with overdue days = 0.

An overview of the columns in the table:

  • ACCOUNTINGDOCUMENT – the invoice number.
  • DOCUMENTDATE – the date of the invoice.
  • AMOUNTINCOCURR – the amount of the invoice in the company currency.
  • NETDUEDAYS – the number of days between the invoice date and the due date. This is also known as the payment terms.
  • OVERDUEDAYS – the number of days the invoice is currently overdue by.
  • PRED_CLEARINGDAYS – the predicted number of days between the invoice date and the clearing date or payment date.
  • PRED_OVERDUEDAYS – the predicted number of days the invoice will be overdue by. This is calculated by subtracting the net due days from the predicted clearing days.

Story Page: Customer Analysis

The second page in the story provides you with a detailed view on individual customers. A holistic 360° view is provided as of November 28, 2018, including the customer’s historical, current, and predicted payment behavior. The default view on this page is set to customer USCU_L09.

 

The visualizations related to current and predicted payment behaviour are the same as those on the previous page, with a filter applied to reflect data for the selected customer.

In addition to this, we have included information about the selected customer’s historical payment behaviour:

Total Cleared Invoices
The total number of invoices cleared to date for this customer is displayed, followed by the total amount of these cleared invoices.

Average Days Beyond Term to Pay
This value represents the average number of days it takes the customer to clear an invoice in relation to the due date. In this example, customer USCU_L09 pays their invoices an average of 66.69 days late.

Payments by Days Late
This visualization groups historical cleared invoices into buckets which represent the number of days the payment was received late. In this example, customer USCU_L09 has paid $106.631 million late by 61 – 90 days and has only paid $6.875 million on time.

Optimizing Collections with Payment Predictions

Taking Corrective Action
Based on the information we have on customer USCU_L09, we can see that they already have $7.111 million overdue, and they are expected to pay us an additional $1.449 million in the next 30 days. However, our predictive model indicates that we should expect this amount 61 – 90 days late. The predicted payment behavior seems consistent with their historical payment behavior.

How can we control the amount of credit risk we have with this customer? Given we have expected overdue amounts on the way, we could focus our collection efforts more aggressively on this customer to reduce the owed amounts in anticipation of future overdues. We could also limit the amount of credit we extend to the customer in the near future and renegotiate stricter payment terms, until the overdue amounts are back to an acceptable amount.

Deprioritizing Low-Risk Customers
Collections Managers have limited resources and cannot spend time chasing every overdue invoice. If they could identify low-risk customers and invoices that would be paid only a few days late, they could spend more time on high-risk customers. Also, time and expenses could be saved by avoiding unnecessary dunning activities.

In this example, customer USCU_S10 represents a low-risk customer:

Historically, this customer has always paid their invoices early or on-time, as indicated by the negative value for average days beyond term to pay. They currently have less than $1,000 overdue which is negligible, even if the amount is more than 91 days overdue. In the next 30 days, we expect $1.139 million in payment from USCU_S10, which we predict will be paid 1 – 15 days late. From the table with detailed receivables information, we can see that many of these invoices are predicted to be paid only a day late.

With this insight, we can draw the conclusion that this customer is a low-risk customer. Even if their outstanding receivables become overdue and appear on our collections worklist, we know that they will pay shortly after, and therefore it is unnecessary to follow up with the customer or create dunning activities.

Implementing Payment Predictions for Your Organization

This content is delivered for free as part of the SAP Analytics Cloud content library. It is an extension of the existing Accounts Receivable and Collections Management functionality within S/4HANA. For information on how to build this use case with your unique S/4HANA data, please download the latest content library documentation from the SAP Analytics Cloud Help.

SAP Analytics Cloud earns a top ranking from BARC

See how SAP Analytics Cloud performed in the world’s largest survey of Business Intelligence software users.