Why is data categorization fundamental to credit decisioning?

In Credit Scoring, it is not necessary to categorize the data in order to obtain results that can be used to guide decision-making. In fact, an approach based on purely financial variables and calculated on uncategorized data can be used to build a credit score. These will, for example, reflect signals such as the evolution of the bank balance, the volatility of expenses, the number of overdraft days, etc.

This purely financial approach does, however, have its limitations. For example, each month, the sum of consumers' income and expenditure is often equal to or close to zero. However, this does not always mean that a person's budget is limited. An unspent amount may simply be saved, and without categorization it is difficult to distinguish between consumption and saving. Finally, the purely financial interpretation of Open Banking data limits the finesse of risk profiles, as it does not provide a detailed understanding of the borrower's banking behavior. Its reasoning will remain purely accounting.

To take risk analysis a step further, categorization is used to give meaning to financial flows. At Algoan, we capitalize on Open Banking data to provide one of the most powerful risk analyses on the market. Thanks to this categorization of Open Banking data, our credit risk prediction models are fed with more predictive variables and have learned to build finer risk profiles.

Our algorithms enable us to assign each bank transaction to one of our 90 categories, with an accuracy of almost 99%. This granularity of categories enables us to validate 3 major criteria specific to the credit decision and its automation:

• Creditworthiness (affordability):
  Categorization makes it possible to identify incompressible income and expenses (housing, credit, etc.) among all banking transactions. This provides key indicators for solvency analysis: income, debt ratio, living expenses, etc.
• Identification of significant credit events (red flags):
  Categorization enables us to detect applicants who are already in a situation of credit default or financial stress, and thus avoid situations of over-indebtedness.
  
• Creditworthiness:
  Categorization feeds the variables in our credit score, which estimates the probability of an applicant repaying his or her credit installments.

Our categorization allows us to make a 100% open banking credit decision, in compliance with the guidelines of the CNIL's AU-005, which lists the elements that can be used to make a credit decision.

The reliability and complementarity of these indicators enable financial institutions to halve their credit risk compared with traditional categorization and scoring methods. A new era in credit decision-making!

In our next newsletter, we'll take a look at the method we've developed for building categorization and enrichment algorithms for Open Banking data based on Machine Learning. Subscribe to our newsletter on LinkedIn!