Predicted Customer Payment Delay Days help organizations improve short-term and medium-term cash planning by anticipating customer late or early payments.
In IFS Cloud, the Payment Delay Prediction model is based on machine learning and is trained using historical customer invoice data. The training dataset can be limited using configurable training parameters to ensure that the model learns from the most relevant and representative data.
Once trained, the model predicts future payment delays or early payments for newly created customer invoices. The model can be retrained at any time to continuously improve prediction accuracy.
To use Predicted Customer Payment Delay Days in Accounts Receivable Analysis and Cash Planning, the machine learning model must be trained and activated.
Model configuration and training are performed in:
Solution Manager / Automation and Optimization / Machine Learning /Trainable Models
Before training the model, the following parameters should be configured:
EXCLUDE_INTERNAL_CUSTOMERS (YES/NO)
Controls whether invoices from internal customers are excluded from the training dataset. Internal customer
invoices often follow accounting or reconciliation processes that differ from commercial payment behaviour.
Including them may distort payment patterns and reduce prediction accuracy.
INVOICES_FROM_DATE (Date Value)
Defines the earliest invoice date to be included in model training. This parameter allows the training dataset to
be limited to the most relevant historical data. It is recommended to include at least the last 24 months invoice
data.
If no date is specified, the model will automatically use paid invoices from the last 48 periods available in the database for training.
Before the model is trained, certain invoices are automatically excluded to ensure that only
relevant and realistic transactional data is used.
For example, invoices with zero amounts and invoices with negative amounts (such as credit notes) are excluded from
the training dataset. By filtering out such records, the model is trained using meaningful commercial invoice data,
improving the reliability of the payment delay predictions. [will later link the IFS.ai document where we have more
information]
After training is completed, the model must be set to Active status to enable predictions.
Model Retraining
The model can be retrained at any time.
It is recommended to retrain the model on a regular basis. The optimal frequency depends on transaction volume and
business dynamics. Regular retraining ensures that the model reflects the most recent payment patterns and
maintains prediction accuracy.
Tokens are consumed during both training and prediction.
To use this functionality, customers, need to be enabled for Use Predicted Payment Delay in Application Base Setup/Enterprise/Customer/Payment. This option controls which customers should be included in the payment delay predictions. If this option is enabled for a customer, the Pre-defined Payment Delay is overridden from the Predicted Payment Delay.
To enable this option, ensure the Payment Delay Days Prediction model under Solution Manager / Automation and Optimization / Machine Learning / Trainable Models is in Trained and Active status.
The customer payment delays are predicted for each installment in invoices when the invoice is posted. Hence the predicted payment delay days are displayed in the Installment and Discount Plan after posting the invoice. This will consume the tokens as the inference calls are made with the NEXUS platform.
Predicted payment delay days are not calculated if the Payment Delay Days Prediction model is not in Trained and Active status in Solution Manager / Automation and Optimization / Machine Learning / Trainable Models, or if there are connectivity issues with the NEXUS platform during invoice posting.
Note that the predicted payment delay days are applicable only to customer Instant Invoices, Project invoices and Customer Order invoices.
The Customer Predicted Payment Delay can either be analyzed in Customer Installment Analysis page, in Multi-Company Customer Invoices with Interest and Fine Analysis page or in IFS Lobby. Since the Payment delay is predicted per customer invoice installment, these analysis pages will provide an overview to query for the payment delay days. Using predicted payment delay will facilitate the collection process within a company and focus on relevant customers to take necessary action.
Use of Predicted Payment Delay for customers is an option in the Cash Planning scenario level. Cash planning transactions can be analyzed and compared with more accurate cash flow dates with the predicted payment delay days. If the “Use Predicted Payment Delay” is enabled in the scenario level, the cash flow date will be adjusted with the predicted delay days to reflect a more probable cash flow.