DataQuest recently developed a credit scoring model for small and medium sized Dutch businesses applying for a short term loan. The model was based on payment data which we retrieved, with permission from the applicant, from an official PSD II aggregator.
Using payment data rather than annual reports brings two major advantages: 1) payment data is up to date while (approved) account statements may be lagging and 2) payment data reflects the truth while annual accounts may reflect opinions. Hence, using payment data opens a new world of possibilities.
Our approach in developing the model was to first collect the (historical) payments via the aggregator. Following this, the payments had to be classified into categories (e.g. office rental payment). Subsequently we defined a long list of potential risk factors including early warning signals, trends, volatilities (e.g. in income) and financial ratio’s. Finaly, we put together the credit scoring model, the outcomes of which are presented in the form of a dynamic dashboard.
The model enables our customer to accept loan applications efficiently and also sensibly in terms of risk taking.