Migration to Databricks Lakehouse in Three Months
About The Client
Fizz is the first Hungarian, bank-led e-commerce marketplace, operated by a subsidiary of OTP Bank, the largest bank in Central and Eastern Europe. Launched as part of OTP’s “beyond banking” strategy, the platform attracted nearly 7 million visitors in its first year and hosts more than 500 merchants offering over 1.5 million products.
Challenge
In 2023, OTP Ecosystem (Fizz) relied on a SQL Server-based data warehouse with Power BI to support reporting through daily batch processing. As the platform grew, so did the demand for more advanced analytics and faster data availability. The limitations of the existing setup became clear, particularly around scalability, flexibility, and the ability to support modern data and AI use cases.
At the same time, Python-based workloads were running in a separate environment, and scaling the platform required increasing infrastructure costs. Fizz needed a more unified solution that could support SQL, Python, and future AI use cases without adding operational complexity.
To address these challenges, Fizz decided to migrate to a Databricks Lakehouse architecture. The goal was to replace the legacy environment quickly, while maintaining reporting continuity and avoiding disruption for end users.
Solution
Following the decision to move to Databricks, Abylon Consulting supported Fizz in executing the migration from its legacy SQL-based data warehouse to a modern Lakehouse platform on Azure. The focus was on delivering a working solution quickly, while ensuring stability and continuity for existing reporting.
Data ingestion was handled using Azure Data Factory, which collected data from source systems and made it available in Databricks. The platform was built using the Medallion Architecture, with data stored in Delta tables across the bronze and silver layers. Raw data was ingested into the bronze layer, then transformed using Databricks workflows and Python-based processing.
A key objective was to keep the implementation simple and efficient. In the gold layer, business logic and reporting objects were implemented using views, which were directly consumed by Power BI. Existing pipelines and datasets were adjusted and optimized to work seamlessly with the new platform, and Logic Apps were reviewed to ensure reliable data transfers.
Abylon also supported reporting continuity by reconnecting Power BI models to the new environment and optimizing ingestion and reporting performance. To ensure long-term usability, knowledge transfer sessions were held to enable the Fizz team to confidently operate and further develop the platform.
Results
The migration provided Fizz with a modern, scalable data platform on Databricks, delivered in just three months without disrupting existing reporting. The new environment improved performance, simplified operations, and created a foundation for future analytics and AI use cases.
Key achievements:
- Improved performance and scalability without increasing infrastructure costs
- Consolidation of SQL and Python workloads into a single platform
- End-to-end nightly data processing reduced from 3–4 hours to around 90 minutes
- Significantly faster report availability for business users
- Power BI refresh times reduced by approximately 50%
- Large data exports now complete in minutes
- Migration of around 50 tables and core reporting logic to Databricks
- All achieved while maintaining full operational continuity
Interested in the technical details?
Read the full project story on the official Databricks website, including architecture decisions, performance improvements, and lessons learned from the migration.
Other Case Studies
Project Type
Data Platform Modernization & Databricks Migration
industry
E-Commerce / Retail
Technologies USED
- Databricks Lakehouse
- Databricks SQL
- Microsoft Azure
- Azure Data Factory (ADF)
- Power BI
- Python
- Delta Lake
