Scoring

Case


If the company sells goods or provides services with obligations, it is profitable and reasonable to deal only with reliable clients with lowest unpaid risk. The way to predict reliable clients and decrease potential risk of debt is to score the clients.  

Solution


Implementation of scoring model, which helped identify reliable clients and significantly decrease unpaid ratio in the company.
The output of the model was a probability of being a debtor in a shortcoming perspective. Mathematical model was developed using Python.

Challenges


Reprocessing the data of old local scoring model, its extension and merging with new preprocessed data
Local database capabilities restrictions during implementation process
Performance problems during the creation of online updated visual presentation of data using BI dashboards
Harsh deadline

Project info

Industry
Telecommunication
Case
Identify reliable clients and probability of client’s default
Technologies
PL/SQL, MS SQL Server, Python, Microstrategy
Team
DB Developer, 2 Data Analysts, Project Manager.
Terms
2 months for complete implemented model