How Sunweb Accelerated Data Modeling with Agentic AI

The Client

Founded in 1991 in the Netherlands, Sunweb Group is a leading European travel group with an asset-light business model, welcoming around 1.3 million travelers each year and generating approximately €1.2 billion in annual revenue.

Sunweb Group combines the best of both worlds: the convenience and scale of an Online Travel Agency with the expertise and service of a tour operator. By leveraging proprietary technology, data, and deep destination knowledge, the group delivers personalized holiday experiences within reach of the many, offering a pause from everyday life, enhancing wellbeing, and creating lifetime memories. Around 40% of customers return, reflecting strong loyalty and satisfaction.

The group is Europe's largest ski holiday provider, selling winter sports holidays across seven markets on the continent. In summer travel, Sunweb Group is a top-three holiday provider in the Benelux and Nordic markets, with sun holidays focused on destinations across the Mediterranean, sold in five core markets. Besides the brand Sunweb, the group operates Eliza was here, a boutique travel brand offering carefully selected, small-scale accommodations, active in the Netherlands and Flanders.

Sunweb Group employs around 450 people across seven hubs in Europe, with its headquarters in Rotterdam. In addition to its central teams, the group works with experienced employees at destinations, ensuring on-the-ground expertise and high-quality service for travellers. The group was acquired in 2019 by Swedish investment firm Triton.

The Challenge: A Tight Migration Window with No Documented Foundation

Sunweb Group needed to migrate its legacy Customer Data Platform (CDP) systems onto a new Customer 360 platform on Databricks on a tight timeline. They also used that migration as an opportunity to unlock marketing capabilities that the current platform could not support. The exploration surface alone was formidable: over 2,400 SQL objects accumulated across years of departmental growth, with business meaning scattered across metadata, stored procedures, cryptic codes, and tribal knowledge rather than documentation. Reviewing that estate manually would have consumed most of the available runway before any new design work could start. 

To do the exploration thoroughly, Sunweb contacted Xebia.  

The Solution: Using AI to Externalize the Conceptual Layer, Fast

Sunweb teams partnered with Xebia consultants to use Claude Code in line with the client's AI policy (metadata and SQL structure only, never customer data) to perform the archaeological work that the team had no time to do manually. In just two hours, Claude Code parsed the 2,400 objects and produced the conceptual layer the team had never had: a knowledge graph of entities and relationships, a taxonomy of how concepts were named and classified, an ontology of the business rules encoded in stored-procedure logic, a definitions glossary, and a map of the procedural estate. That foundation became the input for stakeholder validation rather than a substitute for it.

"Xebia’s contribution has been instrumental in shaping our new CDP architecture, combining deep cross-technology expertise with a pragmatic approach to selecting the right solutions. They accelerated decision-making and execution, while providing strong project steering and upskilling our team through close collaboration. A great balance of professionalism, expertise, flexibility, and clear communication."

Oriol Pueyo

Head of Engineering, Sunweb Group

The Result: From Knowledge Graph to dbt Star Schema in Days

With the knowledge graph in hand, Xebia’s team ran focused feedback sessions with marketing, data, and architecture stakeholders, iterating a new Kimball star schema in the data modeling logical layer that reflected both current reality and the features Sunweb wants to build next. With the conceptual and logical layers locked, the physical implementation in dbt became close to mechanical. In the meantime, the agents kept the documentation and artifacts in sync as the model evolved, flagging anti-patterns and keeping best practices in check. The result held up brilliantly under adversarial stakeholder review. The solid foundation for the modeling conceptual layer paved the path towards consistent data modeling in Sunweb’s brand-new platform.

If you're wondering what else agentic AI can do for your data platform migration, feel free to contact us.