The evolution of relationship management and creditworthiness assessment models

CRIF, recently accredited as an AISP in 22 European countries, has embarked on a path of significant evolution, positioning itself as digital aggregator for the supply of value added services, and making the most of the opportunities in a digital, regulatory and technological context. In this scenario, optimizing the use of an increasingly broader set of information, including in-house CRIF data, was the basis for the construction of a comprehensive offering of value added services.

The premise behind the logic and the current path is that if, on the one hand, the expansion of the dataset is without doubt a priority, as well as an unmissable opportunity for all the different stakeholders, on the other hand, the pursuit of this starts from the setting of precise objectives (e.g. improvement of the behavioral score,...) as well as a rigorous process of selecting the information that may potentially be useful (also considering the considerable amount of data available today and the tools for processing them, including Artificial Intelligence and Machine Learning).

Using its considerable experience gained in the development of creditworthiness assessment models, CRIF created N.E.O.S. (CRIF New Evaluation Open Suite), a service developed with the dual goal of strengthening the creditworthiness assessment models on the one hand, and on the other, taking advantage of the information made available thanks to the new PSD2 regulation regarding access to current accounts; this drives new approaches to development and customer relations (New Relationship Management).

In the first case, CRIF developed a score based not only on traditional information (credit bureau, in-house data and business information for corporate targets), but also on new information made accessible through the new PSD2 regulation (thanks to the possibility of operating as an AISP).

By correlating current account transactions and traditional assessments (loan payment information, financial statements, other business information...), the new models are shown to be more robust in “predicting” creditworthiness. In this regard, recent analyses, conducted by applying the scoring model to a group of representative banks, have demonstrated that the new score enables 10% higher performance indexes for the Retail segment and 21% higher for the Corporate segment compared to the same indexes calculated using traditional models.

Furthermore, the proposed models increase the breadth and depth of the available information: the increased details on the characteristics of transactions enables, for example, analysis of the cash flow and liquidity passing through business accounts, and estimation of the potential income of individuals.

Analysis of the nature of transactions, together with traditional assessments (risk, business information, credit bureau, ...), is a further powerful information element, not just in terms of prediction, but also for an in-depth examination and for the improvement of customer relations.

In the case of consumers, by studying the transactions, including for example the level of banking with multiple institutions, the approach to saving and the product categories purchased, the opportunities for cross-selling and upselling increase, including across different sectors.

For businesses, analysis of the transactions, together with other information within specific CRIF datasets (e.g. Factoring Index, Cash Invoice Index,…), improves the ability (above all for credit institutions) to identify and meet customer needs, providing the best service possible (e.g. improvement of factoring services and invoice discounting).

Refinement of the study of transactions required the development of an ad-hoc tool, the categorizer. Starting from the transactions of consumer and business customers, this enables an extremely precise identification of the most frequent product categories (up to now more than 50 have been identified). 

The product clusters identified (e.g. Food, Entertainment, Travel, Education, Medical,…) increase the knowledge of customer targets, above all for business development purposes, both for those who hold information (typically banks and financial intermediaries) and potentially for all operators in other sectors which operate around these transactions (the ecosystem is made up of all the players involved in the transaction: industries, insurance companies, utilities,...).

The output from the categorizer enables them to create KPIs which allow the identification of recurring behavior for both consumer and business customers according to certain macro-clusters: the transaction profile (e.g. affluent, business activity,...), the macro-category of interest (e.g. travel, e-commerce,...), product type (e.g. financial, insurance, travel product,...) and spending behavior (e.g. volume, frequency,...).

"Expansion toward new value added services able to take advantage of all areas of potential, including the “broader” datasets within the new context of technological and regulatory opportunities, is certainly a priority. CRIF is investing in this area with the aim of creating and driving new approaches to Relationship Management. However, within the mass of data that can be used, it is crucial to define objectives in advance and clearly identify the datasets that are actually useful, and able to optimize the strong predictive power and customer knowledge through the combined use of the potential of Artificial Intelligence, traditional information (e.g. credit bureau, business information) and expanded data (e.g. account transactions)" - commented Giorgio Costantino, Executive Director, CRIF.