In today's uncertain context, businesses increasingly need reliable tools to solidify their operations, particularly in a scenario where generalized distrust is leading to increased complexity in accessing credit and prolonged payment times. It is precisely in these cases that relying on validated technological tools becomes crucial.
And that's where Nowcasting comes in.
We have already delved into the correlation and importance of the Nowcasting tool for credit risk management in businesses. We will therefore focus more on its validation.
"How do I know that Nowcasting is a reliable tool?"
We are certain that's a popular question when approaching our tool. And we agree to demonstrate its validity not only with standard examples, but with a critical analysis of the results, compared to reality. Until now, even if a model considered so "innovative" did not have enough supporting data, its adoption has been largely popular.
This validation process is aimed at verifying whether the developed model corresponds conceptually and concretely to the intended tool. It is a matter of giving an answer to questions such as: were the parameters set correctly? Are the assigned weights corresponding to the real case? Is a simple analysis of this type enough to correctly describe the problem and provide a solution?
And so, thanks to the validation, we can testify concretely to its actual predictability and reliability compared to traditional data and reality itself.
Validation process
The process involved a sampling of over 260,000 companies, which, as of December 31, 2023, were considered:
- Active: 248141
- In default: 13578, i.e. companies that presented protests, concordatory procedures, or bankruptcy
For the companies designated as in default, the Nowcasting index was taken at 6 months prior to the date of the negative event.
The validation process was carried out using the ROC curve (Receiver Operating Characteristic), which is a statistical analysis used to evaluate the accuracy of predictions made by a financial model. The area under the curve, known as the AUC (Area Under the Curve), represents the probability that the predictions are correct.
A ROC curve with an AUC close to 1 indicates that the model is able to simulate the outcome with high precision and is extremely similar to the real case. For Nowcasting, the determined AUC value is 0.78.
The next step was the comparison between the Nowcasting projection and the MORE score: two scenarios were considered, Best and Worst, selecting:
Best:
- Companies in default that had their last balance sheet filed as of December 31, 2022, a negative event during 2023 and the first quarter of 2024, and the last MORE score between AAA and BB (high-quality values)
- Active companies as of March 31, 2024, selected randomly
Worst:
- Companies in default that had their last balance sheet filed as of December 31, 2022, a negative event during 2023 and the first quarter of 2024, and the last MORE score between B and D (low-quality values)
- Active companies as of March 31, 2024, selected randomly
Regarding the Best scenario, you are able to note that -starting from COVID- the companies that will go into default from 2023 onwards, begin to show a significantly different trend in the Nowcasting indicator, and generally lower than active companies.