Question:
job experience
Answer:
I am working in a bank in the Republic of Mongolia for about ten years and I am focused on uh digital risk management. I created scoring models which gives results and also I made credit policies and was involved in a lot of process of risk management which was built in marketing.
Question:
Tell me about scoring models
Answer:
It's not so easy to create a good scores model because if I mean everyone's situation it's lack of data or lack of good data. You have to uh to make it clear, you have to to underline the aspects which is for you will be you can use and and uh which you can implement because it's you have to understand the structure of this uh data also because a lot of that it's rubbish, you have to to underline the real parameters which uh which are are important credit of customer.
Question:
Given your experience with data quality issues, how do you balance the trade-off between model accuracy and the need for explainability when presenting a new credit risk scoring model to non-technical stakeholders?
Answer:
I, I am trying to to speak to non-technical stakeholders on their language eh to to introduce the scoring model parameters in each eh in their understanding of eh of this life. Eh, I try to to find the ways to explain that this model it's good to give more practical examples. And, and, eh, but eh for me, it's very important to understand that eh the model can be accuracy. I think the accuracy is eh on first step. Eh if I understand this accuracy eh, it's eh the prewar, then, it's um it's uh to to explain to the people who who didn't eh have a technical analysis, you have to be very confident to that they, to understand they that you are expecting this, and eh, they can relay you.
Question:
Beyond model accuracy, how do you handle a situation where a high-performing scoring model results in a significant increase in loan rejection rates for a specific customer segment, and how do you communicate the potential business impact to the product team?
Answer:
I think in in every situation you can you can find the metrics which are understandable and are fair and for business and for risk. If I I understand that this is credit model gives good result and it it can be it can be understood It can be it can be used in the right way and it's a mistake to to give more more credit than cut-off. I will explain it on the concrete numbers on amounts of sums that not only that the company, the bank has not only only profit but and losses from credit. And if on on on this segment on this segment losses are higher than then money which but and achieve it is not so good and it's not the good decision from from point of business.
Better English:
I believe that for any situation, you can identify metrics that are fair and understandable for both business and risk stakeholders. If a credit model performs well and is clearly understood, it can be applied effectively. However, it is a mistake to extend credit beyond the established cut-off. I would explain this using concrete figures to demonstrate that a bank must consider both potential profits and credit losses. If losses in a specific segment exceed the revenue generated, that segment is not profitable, and extending credit there would be a poor business decision.
Question:
When a model's performance begins to degrade due to shifting economic conditions or 'data drift,' what specific steps do you take to identify the root cause, and how do you decide between recalibrating the existing model versus rebuilding it entirely?
Answer:
I am a person who always think that you can do something better and I am thinking about the new model maybe in some in some analysis analysis quality I will will use another model maybe from another point of view and I am a person who have always second and I think if it will be the same situation firstly I will think about about finding the right right right the right course of this situation and and to work on it I think to understand if you want to understand the problem, you have to understand when when this problem begins and if I know where the problem begins, I can make the right decision. I can take the right decision. Decision and I will I will understand what to do better in this situation. Firstly, I will think I will think and I implement um implementation of more
Better English:
I am constantly looking for ways to improve my work. When performing model analysis, I often consider alternative approaches or different perspectives to ensure the highest quality. If I encounter a recurring issue, my first step is to identify the root cause. I believe that to solve a problem effectively, one must understand exactly when and how it originated. Once I have identified the source, I can make an informed decision and determine the best course of action. I am committed to continuous improvement and the implementation of more robust solutions.