B2B Credit Scoring Best Practices

If you are a business with hundreds or thousands of customers, you should consider automated credit decisions and monitoring for your accounts receivable, credit, and collection management. The benefits include faster and better credit decisions, more focused A/R collection operations, better customer service, more effective compliance, and significantly reduced overhead. Employing credit scoring is credit management “Best Practice” when you have a large number of customers and credit applications.

Carixa Suite order-to-cash software can incorporate advanced scoring results into credit management workflow. Here are some benefits to consider from credit scoring, including leveraging your historical data, credit bureau business information, and industry trade payments.

6  Good Reasons to Use Credit and Payment Scoring

1. Speed of Credit Decisions. Scoring can dramatically shorten the time it takes to approve orders, a major customer service and sales benefit.
2. Personnel and Overhead Savings. Scoring automates the decision process, dramatically cutting down the personnel costs associated with credit approvals and letting you do more with less.
3. Credit Policy. You can use credit and payment score ranges to establish corporate policies for acceptable risk and slow payment tolerance. Scoring ensures consistency and objectivity in your application of credit policy.
4. Prioritized Collections.  Use blended credit risk scores to set collection priorities, ensuring the accounts with the highest risk for non-payment get collection attention first. If you focus on “risk,” not just age and value, you will have better outcomes and less bad debt.
5. Customer Advisory. The credit manager can become a partner to sales if you counsel customers on how they can improve their scores by highlighting areas of weakness.
6. Fewer Bad Debts. You can expect reduced bad debts using a valid scoring methodology since many smaller customers would not get the same level of manual review as the larger exposures.

Data Elements Used for Credit Scoring Smaller Customers

Financial metrics for large corporations can be predictive in the 2-5-year range, and many credit analysts still use the traditional or modified Altman-Z Score for this purpose. However, financials are often not available for smaller customers, or they are often stale, unreliable, and subject to fast swings. For smaller buyers, other data elements become critical, many of which can be included in your customer credit application.

1. Years in business
2. Experience of owners
3. Owner’s personal credit ratings
4. Number of employees
5. Annual sales
6. References
7. Business and Industry Trends
8. Financial information (if available)
9. Public records show history, liens, filings
10. Supplier payment experiences 
11. Bank and Lender Experience
12. Credit Line Amount needed
13. Credit exposures of other industry suppliers
14. Reputation

Using Big Data and Machine Learning

Building predictive credit scoring is now significantly easier with the ability to capture data from multiple sources and analyze it with powerful software, where the “machine” can learn from experience and adjust its conclusions. That is, it makes and corrects its decisions based on experience, patterns, and trends in the data. This is called “Machine Learning,” where computers are taught to detect data patterns to predict and validate outcomes with regression testing of past events and then adjust those based on current events. Reaching this goal has gotten much easier due to the rapidly increasing power of computer processing.

Our experience is that leveraging machine learning can significantly improve predictive credit and payment scoring. We use these techniques to build and customize automated credit scores and lines for large trade creditors needing to improve and accelerate credit decisions. As we are a cloud-based service,  there is no software to buy, and we integrate easily with client processes and systems.

Designing a Scoring Model

Credit Scoring Framework has a Simple Flow

1. We agree on the outcomes you want to predict (e.g., bankruptcy, default, severe delinquency, or X days late).
2. The modeling team does all the work in collaboration with you.
3. A training sandbox using data attributes from multiple sources, including your own experience; for example, business standing, financials, and debtor payment histories, often as many as a dozen separate data elements, even unstructured credit data, such as industry “attitudes.”
4. The model can be adjusted for the outcome you aim for and can test multiple outcomes based on “model training sets” created in your sandbox, adjusting the importance or weight of certain elements.
5. More elements do not always better produce a better outcome; picking and testing for the right elements is important. Factors such as years in business, number of employees, and social reputation are critical for small customers. Economic factors are important. If the consumer economy turns down, it is a leading indicator of problems with payments and defaults in subsequent quarters.
6. Machine regression picks and weight-adjusts the critical credit attributes for the outcome you are trying to predict.
7. Using real-time industry data ensures continuous updates to scoring attributes and weights.

Advantages: The scoring models self-adjust continuously without manual updates, pulling in more data types than traditional models, including micro and macroeconomic variables, to target the prediction of outcomes that meet your company’s needs. The models adjust for specific industry or business needs based on your unique data or information, precisely defining your preferred outcome.

Producing Customized Calculated Credit Lines

Calculated credit lines can be integrated with any financial or ERP you use to complete a fully automated process.

• By applying your corporate policies, we can calculate a Dollar Credit Line, taking into consideration a number of factors, including tolerance for risk (are you in a fast growth or more risk-averse environment), lender or insurance limits, or product profit margins (a manufacturer with a 60% gross margin will tolerate more risk than a distributor with a 17% margin).
• Modelers interpret your process and policy, replicating your rules in a computer model. By way of example, where no financials are available, and the customer has fewer than ten employees, you may decide “do not calculate a credit line” but instead perform a manual review or adjust a CCL based on the presence or absence or magnitude of certain elements.
• The algorithms are adjusted through testing feedback to align your scores with your desired outcomes.

Advantages: Custom Calculated Credit Lines are built to your circumstances and adjusted as required by your rules, policies, and both internal and external events. Machine-generated credit lines are useful for accelerated decisions and risk analysis across a portfolio.

Summary

By using computational power and a scientific approach to data analysis, you can obtain extraordinary results for automated credit decision processing and, in doing so, improve credit decisions and customer service while streamlining credit operations.