Colendi project is a credit scoring system that can leverage new sources of information about borrowers to assess creditworthiness in addition to traditional measures. This technology creates financing opportunities to new markets of underserved individuals and businesses. This is accomplished by evaluating complementary and distributed data segments of a user with the help of machine learning based credit scoring technologies. Early in the project, it was already apparent that there existed common data paradigms for storage and evaluation of data sources, which could be implemented across seemingly disparate data streams sharing specific underlying properties.
For example, transaction histories are very valuable tools in assessing creditworthiness, as steady purchase habits over time may reveal stability and regularity that imply higher probabilities of on-time payments and therefore lower probabilities of default. Transactions can be leveraged to score credit across sectors and industries, from simple hardware stores to global shipping companies. They share certain common attributes such as date of purchase, amount, descriptions of items, and involved parties.
However, the same exact model for scoring credit based on hardware store transactions cannot be directly applied to global shipping companies, as the industries and transaction characteristics are very different. The financial sector is fraught with the overextension of models that do not assume an inclusive purpose, thus eventually falling to faulty assumptions. Therefore, it is important to allow the parameterization of any transaction history module to accommodate the major modes of its related credit evaluation across several industries and sectors. The fine-tuning of these base models and necessary supporting data integrations require substantial efforts, expertise, and resources.