It isn’t mandatory to employ data classification to help make decisions around credit scoring. Approaches using purely financial variables calculated from non-classified data can indeed produce credit scores. These include oversight of bank balance activity, expenditure volatility, the number of days in overdraft, etc.

But an entirely financial method has its limits. For example, consumers’ monthly deposits and expenses are often equal to or close to zero. But this doesn’t necessarily mean an individual has a limited budget. An unspent amount can easily be put into savings and a lack of data classification makes it tricky to distinguish between expenditure and savings. In addition, a purely financial approach to interpreting open banking data impacts the accuracy of risk profiles as the banking behavior of applicants cannot be understood in detail. 

Data classification takes risk analysis to new levels by making sense of financial flows. At Algoan, we use open banking data to produce some of the most powerful risk analyses on the market. Open banking data classification enhances our credit risk prediction models by providing more predictive variables trained to develop more accurate risk profiles.

Our algorithms assign every bank transaction to one of 90 categories with around 99% accuracy. The granularity of these categories means three main credit-decisioning criteria can be checked and automated:

  • Applicant affordability
    Classification identifies fixed income and expenditure (accommodation, loans, etc.) This information indicates the applicant’s solvency – income, indebtedness, disposable income, etc.
  • Identification of credit red flags events:
    Classification detects applicants who are already in default or who are under financial strain, thus avoiding situations of overindebtedness.
  • Creditworthiness
    Data classification feeds into our credit score variables to give an indication of the likelihood an applicant will repay any credit. 

Data classification allows us to make 100% open banking decisions that observe the AU-005 directives of the CNIL (the French Data Protection Authority), which list the elements that can be used for credit-decisioning.

The reliability and completeness of these indicators mean financial institutions can halve their credit risk compared to traditional classification and scoring methods. A new era for credit-decisioning has indeed dawned!

In our next newsletter, we’ll look at how we’ve developed classification and enrichment algorithms that can be used with open banking data based on machine learning. Why not subscribe on LinkedIn?