Project Details


Credit Scoring for Credit-Offering Retail Chain

This credit risk assessment task comes from a private label credit card operation of a major retail chain.

The company has been operating its private label card for over 10 years and has applied two different methods for risk assessment with the application's acceptance rate varying from 50% to 75% within this period.

Each accepted application turns the applicant into a client and gives him/her the access to credit for purchasing on the retail chain to be billed 10 to 40 days after the purchase, on a monthly basis on a fixed month day. After his/her credit acceptance, a client would take some time to make their first purchase and receive their first bill. During the first year of using the card, the set of monthly bills and payment behavior is collected and used for credit risk assessment. If the client had any monthly defaults (delays longer than the agreed payment periods) he is labeled as bad, otherwise as good client.

The goal was to exploit the information that was available when the applicant applied for credit and try to predict whether he would be a good or bad client. To achieve this we have used the training data consisting of various kinds of information for the applicants like age, sex, marital status, monthly income etc. and the label (good/bad) to build prediction models. Using the risk models we were able to classify new unknown applicants into good or bad credit score. To verify the credit-scoring model we have used another set of labeled applications from another year. Using this approach we were able to confirm that the credit-scoring model reflects the real data. This resulted in deploying the credit-scoring model in production and consequently was used for selection of new applicants. The direct benefit was in increasing the profits of the company because many applicants would have been otherwise rejected if they were selected manually. Now these applicants are being granted credit and later on they confirm to be good clients. On the other hand, we have minimized the number of clients that were granted credit, but which later on turned out to be bad clients.


We can make a difference together


Let's Talk