Use of Hostinger dbt (data build tool): Enhancing data accuracy and providing insights for growth

Our Data Platform Team participated in Coalesce 2024, Las Vegas, the premier conference for analytics engineers. Coalesce brings together data practitioners to sharpen their skills, connect with peers, and get inspired about the future of data. Our Analytics Engineering Lead, Valentinas Mitalauskas, and Analytics Engineer, Augustinas Karvelis, participated in this event and presented insights and highlights on how we use dbt (data build tool) to improve our data accuracy and drive growth. 

In their presentation, titled Smart Spending at Scale: Implementing Real-Time Slowly Changing Dimensions (SCD) Type 2 with dbt, Valentinas and Augustinas shared how Hostinger leverages dbt to implement Slowly Changing Dimensions (SCD). They also discussed how we used the Debezium connector to overcome the challenges of Change Data Capture (CDC) and how leveraging dynamic incremental predicates allowed us to scale our solution at a fraction of the cost using BigQuery.

The session drew a full house of attendees and sparked engaging discussions throughout the event. Here is the wrap-up for those who couldn’t attend the live presentation. 

Hostinger has experienced tremendous growth in recent years, earning the title of the world’s fastest-growing web hosting company. This is driven by a customer-obsessed approach, with a mission to make online success possible for anyone – from developers to aspiring bloggers and business owners.  

The data team played a crucial role in this, building a robust CDC (Change Data Capture) and STD (Streamlining Data) pipeline to track and analyze customer actions during that critical first day.

The solution was built around three core pillars: reliability, affordability, and scalability. It ensures dependable and consistent data, is cost-effective (avoiding expensive data warehouses), and can scale alongside Hostinger’s rapid growth.

However, the previous data replication process faced several challenges. Issues with tracking hard deletes, capturing only the latest changes instead of the full picture, and discrepancies due to hotfixes made the data unreliable. Additionally, replicating entire tables to fix historical snapshots was inefficient and prone to errors.

The solution that the data team implemented addressed these challenges, improving data reliability, simplifying data models for better accessibility, and ensuring the system could scale as Hostinger continued to grow while keeping costs under control.

Want to dive deeper into this topic? Watch the full presentation and Q&A session from Coalesce 2024 for more insights.