Credit And Risk

Calculating lifetime loss for loan portfolio

IFRS9 was the new regulatory guideline for European banks that they have to comply with by the end of 2017. Other banks in the Middle East and Asia are also accessing their retail portfolio losses spread over the lifetime of the credit to stay ahead of curve.


Lifetime Probability Of Default

(Econometric Model)


Loss Given Default

(Empirical Model)


Exposure At Default

(Empirical Model)

Lifetime Assessment

Lifetime of different credit products was estimated

Stage Allocation

Accounts were segmented based on their riskiness at any given time compared to the time of origin

Performance Testing

All the models were assessed based on certain parameters

The Challenge

Losses were calculated based on the expected performance of the portfolio for the next 12 months, which is a very short duration for a product considering its overall lifetime.

European banks have to follow IFRS 9 guidelines to calculate the Expected Credit Losses. Current models access the PD of accounts for only 12 months which is an under assessment of provisions that banks keep to cover for any credit losses.

Estimating the lifetime of credit cards and current accounts is tough due to the nature of the product.

Model validation was crucial to submit final reports to the management to give confidence in the models. Presentation to executive-level members requires covering all the details in a simple manner.



Historical data was used to access the lifetime and multiple options were provided to the management.

Stage allocation was done on the basis of change in PD of an account since its origin.

State Of Economy Model was designed and used to calibrate the model outcome which was primarily based on the change in House Prices and Unemployment rate of the state.

Increasing competition forcing organizations to give more credits to their existing and new customers. Covid-19 added more complications and uncertainty to the future. Due to which banks and other financial institutions are under immense pressure to design systems that give faster results with rocket science equivalent precisions in decision making. Accessing the creditworthiness of customers is the top priority of risk managers so that the firms don’t end up bankrupt.

Low interest is eating the profit margins. In some countries, it is even negative, which means your savings are going down resulting in increasing credit book as the only solution available. Central banks are also easing borrowing conditions in their respective countries. However, recovery from bad loans is still the bank’s own responsibility.

In this environment, designing an efficient credit scoring system is one of the best investments a credit institute can do. Manage risks by introducing new clients to the portfolio and find out new markets for expansion keeping fraudulent customer identification in mind. Risk scorecards are widely used in banks and financial institutions for multiple purposes like acquiring new customers, maintaining a risk profile of existing customers, calculating expected losses on a portfolio, and finally collections and recovery management. In past, due to the high costs of the scorecard building exercises, these econometric models were only affordable by big companies, however, with the introduction of new open source technologies designing, implementing, and maintaining these models is low complexity and high priority project.

Below points should be considered before the development of any scorecard:

How reliable is the data? Data is the first building block for any decision-making engine. It should be error-free and assumption-free. Always remember garbage in, garbage out.

Assumptions – Calculated risk is the key, speculation is the lock that you won’t be able to unlock.

Existing solutions – Check what are the existing solutions available with the organization. There is no need for redesigning the wheel if a workable solution is already available.

Regulations – Credit is one of the most regulated industries in the world. one should always be updated with the regulations in the industry.

Quick implementation – In this ever-changing world, there is no use in designing solutions that will take years to build. solutions must be built, tested, and implemented as quickly as possible.

Easy Interpretation – Black box solutions are hated by everyone. one should be able to understand and explain the models easily in a simple and lucid manner.

Scorecards should be based on a strong foundation of robust data and well-tested econometric methodologies. Technology should be used to implement new ideas and to make a User Interface that can be used by non-technical people without knowing many details about the complex architecture. The design of the framework should be modular so that changing parameters impacting the model can be easily updated. Dynamic data plays an important role in fine-tuning the performance of the models. For loans with a longer maturity, like mortgages, updating the application level is important. There is no point in using the salary of a customer for a loan that originated 10 years back.

At Trugo Consultancy Services analytics division, we believe that analytics is for everyone who wants to tell their story through data. If a solution can be delivered by simple spreadsheets, we do it. If there is a complex situation where we didn’t know the end results, we explore. We always try to leverage the business knowledge of key stakeholders and try to bridge the gap of organizations through data, analytics, and technology.

The Project

Project type – Predictive Modelling 
Industry – banking
Technology – SAS
Delivery Time – 2 months

Consultants worked – 07