By Yafit Solenko, VP Finance and Elinor Schmorak, CCO at Curve.tech

Businesses have different methods to evaluate potential credit risk, most of which rely on experience-based judgments. Still, in many cases, those methods are time-consuming, subjective, and have limited capability of identifying risk in advance. Credit default prediction is a way for businesses to take better control of the organization’s cash flow, avoid risky transactions, tailor deals in general, and specifically payment terms to minimize credit risk.

So, what is credit default prediction, and how does it work?

Credit default prediction is an AI-based process that is designed to enable businesses to monitor and better evaluate the behavior of their customer’s future payments. In this process, existing customer-related financial transactions are segmented into two groups; The first group consists of transactions that are predicted to be paid according to their credit terms and the second, of transactions that are predicted to be overdue.

For the second group, the forecast module is used to predict the probability of a transaction to eventually be paid and provide a more accurate range of the expected delay in payment.  

Let us review the following case example:

RAN Auto Parts of Brooklyn, NY, has been in business for the past 20 years, with hundreds of customers across North America. The company offers a variety of payment terms (such as +30 Days,+60, +90, Bi-Monthly, etc..) for different customers, some change from one order to another, such as when discounts are being offered to speed up cash flow.

One of the customers in South Carolina has missed their payment due date several times over the past few months. In order to define the risk category that will be applied to this customer, RAN is reviewing its accounts receivable aging report and assess a series of High-Risk Indicators that they will use to determine whether a customer has a “Standard Risk” or a “Higher Risk” status.

The fact that this customer had a bad-paying history was evaluated along with other factors such as; how many invoices were fully paid on time, what was the extent of the delay in past payments, what is the customer’s location (local or foreign), the source of the payment transferred, and are there any changes to laws and regulation which can cause delays in transfers.

Now, if RAN decided that this customer has a “Higher Risk” category, does that mean that RAN should consider ending business relationship with them? Imagine you were in RAN’s shoes, trying to evaluate the viability in signing a new deal with that “High Risk” customer. You could and probably should ask yourself the following questions:

  • Will this customer pay for their next order?
  • Assuming they will- when is the next invoice likely to be paid?
  • What is the best way to decrease the credit risk for this customer? For example:
    • Prepayment-based orders?
    • Increased transactions monitoring and analysis?
    • Increased level of continuous control of the business relationship with the customer?

Once you ask yourself these questions, you are just beginning to see the advantages and capabilities of credit default prediction. Predicting the particular risk derived from this one customer is very likely related to all the factors that were mentioned above and maybe, a few other factors that one can observe and analyze.

With that said, by scanning the business historical data and market trends, we can detect and engineer many additional factors. These factors might not be friendly to the human analyst but can be efficiently analyzed by learning machines. This is the reason why an algorithm can predict the risk even when no past payment was overdue or classify a new client that has no payment history yet. Using automated prediction can help you answer the above questions efficiently and get a higher level of accuracy.

What are those additional factors that provide insight into risky customer’s behavior before a crash occurs? The obvious are the payment terms and the customer payment history, but there are also less obvious factors such as products/services related risks, yearly seasonality, market trends, and much more.

Here are just a few of the primary reason’s businesses could benefit from predicting their customer payment cycles:

  1. Better predicting your future income can help you better plan your future expenses and therefore have better control over your business’ cash flow.
  2. By identifying risks in a timely fashion, you will be alerted as to when invoice payments are predicted to be late or not paid.
  3. Monitoring all your customers automatically in a single view.
  4. Identifying which are the customers who are less trustworthy and act accordingly.
  5. Minimizing credit risk by simulating the optimal deal, based on a range of parameters such as payment terms, product mix, and quantities.

Credit default prediction can monitor all your customers’ information while running it through an ML-based prediction algorithm and guide you in making better-informed decisions when risk is involved. What could take you days is done within a few seconds.

This is the Power of Credit Default Prediction. Discover it for yourself.

Do you want to hear more about how you can use machine learning for credit default prediction? Reach out to us for a consultation at info@curve.tech