Trainline Scales Self-Service Analytics with Starburst Galaxy

About

Trainline is the largest independent rail and coach travel platform in the world. The company aggregates routes, fares, and journey times from around 270 carriers across more than 45 countries and sells 250,000 tickets in a single day across more than 25 countries. Kiron Ponnath works as a Staff Data Engineer in Trainline’s data platform function, where he focuses on building secure, scalable tooling for analytics, data engineering, experimentation, data science, and machine learning teams.

Challenge

Previously, Trainline’s SQL workloads primarily ran on cloud-native engines. While these supported many use cases, they encountered strict timeouts on queries that ran longer than roughly 20–25 minutes. That created problems for data science and analyst teams that wanted to continue using SQL for larger scans and longer-running queries instead of moving those workflows to PySpark or other engines.

Trainline also wanted to improve cost efficiency, reliability, and security as more teams across the business required direct access to data. Historically, onboarding users involved manual role setup and coordination with infrastructure teams, making access management slower and more complex. To simplify governance, Trainline developed an in-house, platform-agnostic access management tool that centrally defines roles and policies, then generates platform-specific permissions, including IAM policies for AWS and SQL grants for Starburst Galaxy. This approach provides a scalable foundation for secure, governed self-service analytics across the organization.

Solution

Trainline adopted Starburst Galaxy to support long-running SQL workloads and provide a more reliable, scalable query engine for its data teams. With fault-tolerant execution, Galaxy enables analysts and data scientists to run large, complex queries without interruption.

At the same time, Starburst Galaxy simplified secure data access through centralized governance and group-based permissions, making it easier to onboard users and expand self-service analytics across the business.

Results

• 20%–30% cost reduction in year one: After migrating larger workloads to Galaxy, Trainline reduced overall costs in the first year while staying within budget.
• Reliable support for long-running SQL workloads: Galaxy addressed the 20–25-minute Athena timeout challenge for analyst and data science teams.
• 300–400 active users: Starburst Galaxy is now used directly by teams across analytics, experimentation, fraud analysis, marketing, finance, and product.
• 2-hour onboarding for fraud analysis: Using Entra ID groups and role-based access, Trainline onboarded the fraud analysis team in just 2 hours, compared with a process that previously took days.
• Simplified governance and access control: Trainline’s Unified Security Platform allows roles and policies to be defined once in a platform-agnostic way and applied consistently across multiple data platforms. For Starburst Galaxy, those policies are translated into SQL grants and synchronized through the Starburst API.
• Strengthened data governance: Combined with SSO, SIEM integration, RBAC, ABAC, and tag-based column masking, Starburst Galaxy enables Trainline to better protect sensitive data while providing governed access for business users.
• Simplified Apache Iceberg maintenance: Data science and analyst teams can run optimization, vacuum, and related maintenance tasks without building custom Airflow jobs from scratch.
• Improved business decision-making: By enabling faster, more reliable access to data, Trainline supports teams such as fraud analysis, marketing, and finance in making more timely and informed decisions across the business.
• Operational visibility: Starburst Galaxy’s telemetry tables support Trainline’s FinOps and access control initiatives by providing greater visibility into platform usage and operations.

With Starburst Galaxy as its core data query engine for heavy SQL workloads, Trainline has expanded secure self-service analytics across the business—enabling teams to make faster, more informed decisions while maintaining the flexibility of its hybrid lakehouse architecture.