Product descriptions:
Banco Bradesco has 80 years of experience in financial services and is one of Latin America’s largest banks, serving over 75 million customers across seven countries. In digital banking, being able to quickly respond to customer interactions is critical for success. However, Banco Bradesco’s legacy technology systems made it difficult and expensive to collect and analyze customer data from digital and marketing channels — and fragmented tools often produced unreliable data. Using Databricks and Lakeflow Declarative Pipelines, Banco Bradesco built a new in-house customer data platform (CDP) that changed how they understand and serve their customers.
The drive for customer centricity
Banco Bradesco’s shift toward becoming more customer-centric placed their customer relationship management (CRM) system at the center of their technology infrastructure. CRM plays a critical role in the bank’s operations — for their main product, credit, the CRM system influenced at least 30% of all transactions in 2024. Supporting customers across multiple channels, from traditional bank branches to digital platforms, the CRM manages communications across 14 main channels and was responsible for 7.7 billion communications in 2024 alone.
The bank’s customer-focused approach generates massive amounts of data. Banco Bradesco processes an average of 2.9 billion raw observations per month from various sources, including mobile apps, push notifications and email communications, plus an additional 2.5 billion raw observations from tags. Managing this enormous data volume fell to a lean team of just nine people overseeing 22 different datasets.
Banco Bradesco’s legacy on-premises systems created significant bottlenecks. The bank often experienced three-day delays in getting data from their CRM, making it nearly impossible to respond quickly to customer needs. “We needed a consolidated view of customer information,” Marcos Abrantes Gomes, senior manager at Bradesco Bank, said. “For example, we might see that 10% of customers interacted with a campaign, but we need timely analytics to determine the best offers for those customers.”
The technical limitations proved frustrating. Data extraction was restricted to one file at a time per channel across 14 channels, and the heavy processing time at the source created even more delays. Without fresh insights into customer interactions and campaign performance, Banco Bradesco struggled to deliver the personalized experiences their customers expected.
Building a modern data platform
To address these challenges, Banco Bradesco turned to Databricks and implemented Lakeflow Declarative Pipelines as the foundation for their new CDP. The bank first used Databricks to ingest raw data in near real time and build tables for their analyst team while ensuring data quality. They built a preprocessing ingestion layer to validate file integrity and auto-request replacements for corrupted inputs.
The solution leverages the medallion lakehouse architecture, ingesting raw data through the Bronze layer, cleaning and validating it in the Silver layer and applying dimensional modeling and aggregation in the Gold layer. Lakeflow Declarative Pipelines defines and executes the full pipeline logic. “We have 14 channels with different data sources, and Lakeflow Declarative Pipelines is responsible for all the process executions running in parallel with one cluster,” Marcos explained. “This allows us to control costs, which is very good for us.”
The data validation ensures only clean data progresses through the pipeline, eliminating the need for reprocessing that plagued the bank’s on-premises system. Instead of reprocessing entire datasets, the new system performs incremental updates with only new data. Since October 2023, the bank has processed 110 billion raw observations — essentially 200 million new pieces of information daily. After implementing Lakeflow Declarative Pipelines, their Gold layer contains 7 billion rows of data, receiving 20 million new rows and updating 15 million daily, plus 40 billion events with 60 million relevant tags processed each day.
The performance improvements have been dramatic. The bank reduced data latency by over 94% compared with their on-premises process, with Lakeflow Declarative Pipelines completing all processes in just 3.5 hours versus the previous 60 to 80 hours. Their main processes now run in three to 40 minutes compared with three to eight hours previously. Perhaps most importantly, the new system requires only two people to operate compared with the eight to 15 people needed for the legacy on-premises process, enabling the bank to reassign those personnel to more strategic tasks.
Transforming customer engagement through data
With their new in-house CDP powered by the Databricks Data Intelligence Platform, Banco Bradesco has transformed their data strategy around customer information. Using Lakeflow Declarative Pipelines, the bank achieved significant improvements in data integrity, reduced latency and processing time and enhanced productivity and business agility. What once required months of reprocessing, weeks of human labor and cumbersome data integrations has been dramatically simplified, delivering substantial operational efficiency gains.
Beyond operational improvements, the platform provides the bank with a robust structure for customer engagement. It now captures and analyzes results from customer interactions to provide feedback to its models, enabling the bank to offer customers the next best action and accurately measure the effectiveness of their marketing campaigns. This closed-loop system creates a continuous cycle of improvement in customer targeting and personalization. “The integration of Lakeflow Declarative Pipelines allows us to trust in our data and to create more contextual offers for customers,” Marcos said. “All the touchpoints customers have in the journey — whether they are spontaneous or triggered by a marketing campaign — are analyzed and transformed into new contextual actions.”
The business impact has been significant. By transforming how they collect, process and act on customer data, Banco Bradesco has transformed their data into a strategic asset — enabling personalized banking experiences at scale and doubling conversion from contextual offers.