MTS Bank reduced time of calculation of assessment of a kredisposobnost three times

2020: Implementation of Fabrique.ai

On July 24, 2020 it became known that PJSC MTS Bank implements in processing system of credit requests a startup of Fabrique.ai from a corporate accelerator of the MTS StartUp Hub center. As banks collect and process Big Data for assessment of solvency (scoring), permanent IT upgrade is necessary for increase in speed of their processing. Solution integration of Fabrique.ai will allow to check more precisely correctness of stages of calculations of solvency of the borrower and to avoid critical errors as the solution automatically analyzes scoring process. The algorithm will increase the speed of execution of separate components of scoring five times and will reduce time for its calculation three times.

The pilot on implementation of Fabrique.ai passed from October, 2019 to February, 2020 within the project of operation and management of lifecycle of models of artificial intelligence and machine learning of MTS of Bank. Based on testing the solution Fabrique.ai allowed to process in peak working hours of the program of consideration of credit requests (the credit pipeline) a flow to 20 requests per second. Thereby the bank could reduce three times time for calculation of models of assessment of solvency.

Fabrique.ai divides an algorithm of calculations of adoption of the credit solution into two components: data collection and calculation of assessment of solvency of the borrower. On the one hand the speed of receiving assessment increases, and with another the solution Fabrique.ai allows to avoid critical errors as scoring settlement process is automatically analyzed.

The pilot allowed bank to receive an alternative view of possible reduction of the general time of response to errors. By means of Fabrique.ai it is possible to reduce time of detection of critical deviations at a scoring calculation stage more than three times. Within testing three mathematical models of assessment were also optimized that led to decrease in average time of their execution five times – from 2nd to 0.4 seconds.

Within the pilot we carried out more granular control of metrics of online monitoring of scoring models of the credit pipeline. By means of Fabrique.ai we managed to reduce time of detection of critical deviations at a scoring calculation stage more than is three times quite notable result which became the main thing of arguments for benefit of implementation. Cooperation with a corporate accelerator once again proved that startups differ in innovation and have the huge IT potential which is necessary for the innovation fintekh, – the head of department of portfolio management and development of scoring cards of MTS of Bank Maria Arinchikhina told.

Thanks to high requirements of MTS of Bank and the selected loading during the project we pumped over skills of a command. We with the Fabrique.ai command in the normal mode will continue to monitor solution integration in a configuration of a scoring system of MTS of Bank, – the founder of Fabrique.ai Sergey Kedrov noted.