Enable customer payment predictions

This article explains how to configure the Customer payment predictions feature on the Finance insights configuration page. This article also includes information that can help you effectively use the feature.

Note

Before you complete the following steps, be sure to complete the prerequisite steps in the Configure for Finance insights article.

Important

If the following data conditions aren't met, the prediction model won't train properly.

To successfully train the prediction model, your environment must contain at least 100 settled customer transactions from the past six to nine months.
These transactions must be distributed across the following three buckets with at least 30 free text invoices, sales order, or customer payment transactions in each:

• On-time
• Late
• Very late

1. Configure the Customer payment insights feature:

   1. Go to Credit and collections > Setup > Finance insights > Customer payment predictions.

   2. On the Finance insights configuration page, on the Customer payment predictions tab, select View the data fields used in the prediction model to open the Data fields for prediction model page. There, you can view the default list of fields that are used to create the artificial intelligence (AI) prediction model for customer payment predictions.

      To use the default list of fields to create the prediction model, close the Data fields for prediction model page, and then, on the Finance insights configuration page, set the Enable feature option to Yes.

   3. Specify the "very late" transaction period to define what the Very late prediction bucket means for your business.

      For each open invoice, the system predicts the probability of payment in three buckets: On time, Late, and Very late.

      • On time – This bucket includes payments that are predicted to be paid on or before the transaction due date.
      • Late – This bucket includes payments that are predicted to be paid after the transaction due date but before the start of the "very late" transaction period.
      • Very late – This bucket includes payments that are predicted to be paid after the start of the "very late" transaction period.

      Note

      If you change the "very late" transaction period and select Change late threshold after the AI prediction model for customer payments has been created, the existing prediction model is deleted, and a new model is created. The new prediction model will move transactions into the "very late" period, based on the settings that were entered to define it.

   4. After you've finished defining the "very late" transaction period, select Create prediction model to create the prediction model. The Prediction model section on the Finance insights configuration page shows the status of the prediction model.

      Note

      At any time while the prediction model is being created, you can select Reset model creation to restart the process.

   The feature has now been configured and is ready to be used.

After the feature has been configured, and the prediction model has been created and is working, the Prediction model section of the Finance insights parameters page shows the accuracy of the model.