Security Bank Corporation (SBC), a leading financial institution in the Philippines, faced several challenges that were hindering operations: inefficient time deposit models for liquidity risk management, manual and time-consuming model deployment, fragmented risk management calculations, and a lack of standardization and automation.
Let’s zoom into three specific use cases to highlight how Dataiku — the Universal AI Platform — has helped the team at SBC address those challenges by optimizing operations and transforming their approach to liquidity risk management and credit loss valuations.
Dataiku streamlined the compliance reporting process, enhancing efficiency and reliability, and boosting the bank’s reputation with regulators and stakeholders.
Ellen Grace Goromeo Data Scientist, SBC
Revamping Time Deposit Models for Liquidity Risk Management
SBC provides time deposits to its customers and, upon reaching maturity, customers have the option to extend the term of their existing time deposit. This process is known as a time deposit rollover. When customers opt to rollover a time deposit deal rather than liquidate it upon maturity, it affects the bank’s liquidity risk profile by modifying the timing of cash flows. Consequently, it’s essential to implement a robust model to effectively manage liquidity risks in the face of changing market conditions.
However, the existing models were built by an external service provider using a limited set of aggregated data, resulting in suboptimal forecast accuracy for time deposit rollovers and inefficient management of liquidity risk. With regulatory pressures mounting, SBC sought a more robust and internally-controlled solution.
SBC turned to Dataiku for its machine learning (ML) capabilities to develop a more accurate, scalable model in-house. With Dataiku, the team shifted from using aggregated data to deal-level data, improving accuracy and model reliability. Dataiku enabled efficient data cleaning, reduction, transformation, EDA, and model development, while fostering collaboration between data scientists, quality testers, and risk validators.
The streamlined process allowed the bank to consolidate its twelve distinct time series models into a single, high-performing model, meeting regulatory standards while significantly reducing maintenance costs and enhancing operational efficiency. With Dataiku’s help, SBC met its deadlines and delivered a solution that improved compliance, reduced operational costs, and optimized liquidity risk management.
Dataiku made it easier to run iterations and explore various ML models simultaneously under different scenarios and parameters, leading to the best model based on desired metrics.
Ellen Grace Goromeo Data Scientist, SBC
Enhancing Financial Stability With Automated ECL Calculations
SBC’s Risk Management Group (RMG) needed a more efficient way to handle the complex task of calculating expected credit losses across multiple portfolios. The calculations, while critical for accurate risk assessment and regulatory compliance, are inherently complex due to the varying nature of each portfolio, the need for high precision and timely calculations, the need to ensure scalability to handle increasing data volumes, and the necessity for continuous collaboration and training to develop and deploy the calculations accurately.
All of these variables made the solution ripe for automation. With Dataiku, the team streamlined and automated the ECL calculation process for 15 portfolios. Dataiku facilitated collaboration between data scientists, risk analysts, and business professionals, enabling real-time scenario analysis and ensuring data governance.
- First, the calculation process of each portfolio was onboarded, which required a thorough understanding of the existing manual ECL calculations. Risk analysts ensured regulatory compliance and alignment with the bank’s risk management objectives. This step ensured all necessary components were accurately captured and ready for integration into Dataiku.
- Next, the data scientists automated multiple project flows while keeping them consistent. It involved creating scenarios that integrated fragmented input data from external sources, performing necessary transformations, and more.
- Portfolios were deployed in the automation node of Dataiku to ensure that the entire calculation process was reliable and repeatable while minimizing the risk of human error. Data scientists also configured the scenarios such that the output is accessible to stakeholders from other divisions.
The platform automated critical processes, significantly reducing the manual effort required for ECL calculations. Dataiku’s integration with existing systems allowed SBC to meet regulatory deadlines ahead of schedule, providing valuable insights faster and with greater accuracy. Automation capabilities reduced manual data preparation and ECL calculation times by 75%, enabling risk analysts to focus on higher-value activities like model optimization and scenario analysis. Plus, the team upskilled over 10 business professionals in data analytics and risk modeling during the use case, further democratizing AI usage across the company.
Dataiku seamlessly integrated into our existing tech stack, allowing us to scale without additional infrastructure investment. Its scalability ensures we can expand as our data grows and models evolve without performance bottlenecks.
Robin Kamille Ramos Data Scientist, SBC
Transforming Model Deployment With Automation in Dataiku
Prior to implementing Dataiku, SBC faced significant delays in deploying ML models for cross-selling strategies. The manual process of running models and generating outputs consumed up to 16 hours every month, diverting resources from model optimization and innovation.
Manual processes also hindered collaboration among team members, as there was no centralized platform for sharing and managing models and data. Without central repositories for notebooks and models, locating and accessing the necessary code became challenging, leading to inefficiencies and delays. The absence of standardized coding practices further contributed to disorganization, making it difficult to understand and maintain models. This lack of structure and consistency resulted in hard-to-maintain models, ultimately impacting the team’s ability to deliver accurate and reliable results.
By migrating its model deployment processes to Dataiku, SBC automated routine tasks, reducing model run-time by 75%. This reduction not only underscores an increase in efficiency, but allows for more rapid iteration and deployment cycles in the future. Dataiku’s automation capabilities allowed models to be deployed faster, improving the efficiency of the cross-sell team and providing more timely insights to business users. The centralization of processes in Dataiku also enhanced collaboration and improved model accuracy by ensuring standardized workflows.
Dataiku’s built-in governance and compliance tools provided full transparency into data sources, version control, and audit trails, ensuring all calculations met stringent regulatory standards. This made the risk management process more robust and auditable, reducing compliance risks.
Robin Kamille Ramos Data Scientist, SBC
Along with streaming the model development process, the team also used Dataiku to develop data quality monitoring metrics and create notifications for each data scientist upon the completion of specific modeling tasks, ensuring up-to-date data quality monitoring. The significant reduction in model run-time allowed SBC’s data scientists to focus on high-value tasks, including model refinement and new feature development, while providing quicker insights to the business for more agile decision-making.
The Impact Across SBC’s Operations
The integration of Dataiku into SBC’s operations has yielded transformative results across three key areas: liquidity risk management, credit loss calculations, and ML model deployment. By automating processes, enhancing collaboration, and improving data governance, SBC has realized significant gains in efficiency, cost savings, and regulatory compliance.
The bank is now better equipped to innovate, enhance customer acquisition, and maintain its competitive edge in the financial services industry.
The automation provided by Dataiku allowed us to deliver model results faster and more efficiently to our business partners. By centralizing our processes and automating routine tasks, we not only improved productivity but also enhanced collaboration among team members. The platform’s capabilities enabled us to maintain high standards of accuracy and reliability in our models, leading to better business outcomes.
Patricia Manasan Head of Data Science and VP, SBC
