Ushering personalized banking for millions of customers

Complex data silos slow the delivery of data-driven solutions in banking

Techcombank, the largest private financial institution in Vietnam, is spearheading the digital transformation of the financial sector. With a mission to enable individuals, businesses and corporations to progress and thrive sustainably, Techcombank leverages data as a key competitive advantage. The bank operates a vast network of 315 branches serving over 15.3 million customers — a testament to their significant influence on Vietnam’s financial landscape. Their strategic initiatives underscore their commitment to leveraging data to drive customer-centric innovations and operational efficiencies, effectively expanding their financial network.

The bank’s data landscape was characterized by a multitude of disparate systems, including numerous on-premises databases and a legacy data warehouse, which created a steep learning curve and hindered collaboration among teams. Additionally, the massive volume of transaction data added complexity to the infrastructure and slowed down decision-making processes.

Governance was another significant challenge, as users were generating their own datasets in various areas, complicating access and security management. The development experience on legacy platforms was also problematic; data teams often had to manually manipulate data in Excel to resolve discrepancies, which was time-consuming and error-prone.

Recognizing these bottlenecks, Techcombank initiated a comprehensive transformation to onboard the Databricks Data Intelligence Platform, unifying intelligence from various systems into a cohesive data ecosystem. This strategic move was made with the ambition to empower users, enhance data analytics capabilities and enable the bank to harness the full potential of data, ultimately fostering a culture of data-driven decision-making throughout the organization.

Central to the initiative is leveraging enhanced AI models and analytics for a comprehensive customer understanding. Dubbed the “Customer Brain,” this powerful customer 360 tool aims to centralize all customer data within the bank’s extensive network, providing extensive insights that are pivotal for crafting targeted marketing strategies and personalized product offerings. By understanding customer behaviors and preferences at a granular level, Techcombank can offer precisely timed and contextually relevant solutions, enhancing customer satisfaction and engagement. The Customer Brain also acts as the omnichannel orchestrator by feeding intelligence into all sales and customer touchpoints.

Another significant innovation is the Lead Allocation Curated Engine (LACE). This sophisticated system revolutionizes how leads are managed within the bank by aggregating detailed customer profiles and behaviors leveraged on the “Customer Brain” initiative. LACE equips relationship managers with enriched data and actionable recommendations, ensuring that leads are not only intelligently assigned but also effectively engaged. This tailored approach helps in optimizing the allocation of leads, and increasing conversion rates by ensuring that customers receive attention from the most appropriate and informed personnel. “When it comes to lead allocation, it’s crucial to optimize the distribution among team members to enhance conversion rates while managing a significant workload,” Tran Viet Hung, Head of RBG Business Development at Techcombank, explained. “This strategy not only maximizes lead utilization but also improves overall conversion rates. The AI insights from LACE deliver comprehensive analyses of financial situations and customer spending habits, along with the latest triggers for engagement, significantly reducing the time spent searching for customer information. Moreover, the AI scoring provided by LACE enables sales teams to effectively prioritize leads, thereby boosting conversion rates.”

Techcombank has implemented their Geosense product to strategically expand their merchant network. Geosense assists frontline workers by providing them with analytics-driven insights to target the right merchants to approach and the optimal strategies for engagement and upselling. This tool is crucial for broadening the country’s financial inclusion, particularly by targeting businesses and merchants who are under-banked, ultimately providing an economic uplift.

The engineering team also built out solutions for their fraud prevention and risk management use cases by leveraging Databricks. They implemented machine learning (ML) models to detect and mitigate fraudulent activities across their product lines, safeguarding both the bank and their customers. They also developed robust models for credit risk management, aiming to reduce the incidence of credit application fraud. These models help in assessing the creditworthiness of applicants more accurately, thereby minimizing financial risks and ensuring the sustainability of lending practices.

Empowering users with Databricks SQL, Notebooks and AI for data democratization, innovation and transformation

Techcombank’s utilization of the Databricks Data Intelligence Platform spans various components, each serving a strategic function in the bank’s data ecosystem.

Databricks Notebooks have significantly empowered Techcombank’s teams, unlocking the full potential of their data and transforming collaboration and productivity across the organization. With a user-friendly interface, these notebooks facilitate seamless teamwork among thousands of active users, enabling them to share insights and work on projects concurrently. This collaborative environment enhances efficiency and drives innovation, allowing teams to respond more swiftly to business needs.

Moreover, Databricks SQL has democratized exploratory data analysis (EDA) across all departments, simplifying ad hoc reporting and enabling rapid, informed decision-making. By handling countless queries daily, this tool empowers users to confidently leverage data, fostering a culture of trust and data-driven insights throughout the bank.

The scheduling function within Databricks has also played a pivotal role in automating a significant portion of repetitive tasks, freeing up valuable time for users to focus on higher-value activities. This automation streamlines workflows and enhances overall productivity, allowing teams to concentrate on strategic initiatives rather than mundane processes.

The Data & Analytics team has received positive feedback from various departments across the bank, highlighting how the platform has enabled teams to enhance their daily operations. The flexibility of utilizing Python alongside SQL, combined with automated workflows and SQL alerts, has notably reduced the cognitive load for the finance MIS team. This democratization of data empowers them to meet their service level agreements without requiring after-hours work.

As Quang Canh, Head of Risk Analytics at Techcombank, noted, “With a twentyfold increase in credit applications from retail clients and a threefold increase from corporate clients, Techcombank is managing vast amounts of customer data sourced from over 50 distinct systems. The bank has been utilizing more than 100 machine learning risk models to streamline its credit lifecycle management operations. The integration of Databricks has provided a cohesive platform for data analytics and advanced machine learning, significantly minimizing costs associated with departmental silos and fostering enhanced collaboration across teams. Most notably, Techcombank’s implementation of an enterprise feature store with over 7,500 features has further optimized its machine learning capabilities. This feature store serves as a centralized repository, supplying customer aggregated and contextual information in a timely manner. It not only streamlines the process from model development to implementation but also ensures consistency and governance of data across various applications. The bank can now make more informed decisions in real time, which is critical in the fast-paced financial environment.”

The power of MLOps on Databricks with MLflow capability has fast-tracked model implementation processes that previously took months; now, teams can experiment with and execute models in just weeks. This acceleration allows for quicker iterations and deployments, enhancing the bank’s ability to innovate and respond effectively to market demands. In just a couple of months at the end of 2024, MLOps has powered the release of a handful of pioneering models for Techcombank, ranging from risk management to credit uplifting. As Techcombank advances toward sophisticated machine learning applications, MLflow has become indispensable for managing the lifecycle and production of machine learning models. Even in its early implementation stages, MLflow is enhancing MLOps processes, equipping teams with the capabilities they need to innovate effectively.

Leveraging graph analytics through GraphFrame technology has enabled Techcombank to gain deeper insights into customer connections that were previously obscured by traditional relational databases. This capability allows the bank to understand complex relationships within their data better, driving more informed strategies for customer engagement and service enhancement.

Uplifting the engineering experience, and looking forward to the future

Since implementing the Databricks Data Intelligence Platform, Techcombank has witnessed tangible improvements in their operational efficiencies and data-driven decision-making capabilities. The introduction of self-service functionalities through Databricks has proven highly successful, enhancing productivity and enabling teams to independently access and analyze data, which has considerably expedited operations.

The performance of data pipelines has seen a significant increase, streamlined processes and reduced time to insight for critical banking operations. The initial pilot project on Databricks demonstrated a significant enhancement in the delivery of data pipelines due to simplified integrations, resulting in much quicker ETL processes. These advancements have directly contributed to improved service delivery and heightened customer satisfaction by facilitating faster and more accurate data processing.

Looking ahead, Techcombank is set to transition their data platforms from AWS native services to Databricks within the next year. This migration will also involve shifting from native data formats such as Parquet and Apache Iceberg™ to Delta Lake, enhancing data reliability and performance. In terms of machine learning, the bank aims to centralize a wide range of features to the enterprise feature store and multiple ML models, including traditional machine learning models and large language models, on the Databricks Platform. This centralization will streamline model management and deployment, significantly enhancing predictive analytics capabilities across the bank’s operations while also ensuring adaptability to future technological advancements.

The transition to Delta Lake will mark a significant upgrade from a legacy data platform that relied on a mix of Parquet and Iceberg file formats. Delta Lake has enhanced data recoverability and simplified merging processes. The potential for increased robustness, performance and resiliency allows for data rollbacks, when necessary, through time travel capabilities — transforming ACID transactions and enabling efficient handling of large-scale data, all without locking the bank into a single data format.

Generative AI (GenAI) is a key focus for Techcombank as the bank seeks to enhance customer experiences through intelligent insights and improve operational efficiency. Techcombank is developing an internal chatbot, Smartie, based on Vector Search — a RAG framework powered by Databricks. This dynamic knowledge base will streamline access to essential information for employees, further enhancing operational efficiency. Together, these initiatives position Techcombank at the forefront of AI-driven innovation in Financial Services, equipping users with no-code and low-code analytics capabilities to better understand and leverage data. “On top of LACE, Smartie is another data product that truly enables the team,” Nguyen Thi Thuy Linh, Director of Business Development and Campaign at Techcombank, noted. “Instead of reading through extensive documents, users can easily access concise and accurate answers. This significantly reduces the time spent on information retrieval and gathering.” Additionally, the Data & Analytics team is piloting AI/BI Genie, a conversational tool designed for business teams to engage with their data using natural language. This initiative aims to empower users to analyze critical metrics without needing SQL expertise, enabling them to gain deeper insights into customer behavior and inform strategic decisions. Techcombank is committed to democratizing data access and empowering users across the organization to harness the power of data for enhanced innovation and decision-making. By implementing advanced capabilities such as graph analytics and generative AI, Techcombank enables their employees to engage with data intuitively, without the need for extensive technical expertise. This approach not only fosters a data-driven culture but also enhances the bank’s ability to derive meaningful insights from complex customer relationships.

As a result, Techcombank is positioned to leverage these innovations to create personalized experiences, drive operational efficiencies and maintain a competitive edge in the rapidly evolving financial landscape.