Related Services
dbt Cloud, GCP, Terraform, BigQuery, GitHub, Cloud Build
Industry
Agriculture
Company
Xebia Data
Kramp is a distributor of spare parts for the agriculture and construction industries. Founded in 1951, the family-owned company based in Varsseveld is the largest supplier of spare parts for its industries in Europe. It has over 3,000 employees, B2B annual sales of 1.16 billion euros in 2023, and 12 operation centers. Over the past 73 years, Kramp has always pursued one goal: to make work as easy as possible for all its customers.
In Search of Data Mesh
Kramp had been leveraging its Google Cloud Platform (GCP) since 2021. As the number of projects deployed on the platform grew, it became clear that Kramp needed an efficient tool to perform data transformations and ensure that queries were properly documented and tested.
Since Kramp is a consolidated organization with clear domains (product, marketing, finance, etc.), the company required a data infrastructure that could sustain an independent way of working. A data mesh, a model where all departments can pull and push data to a central location, was the perfect solution.
Kramp wanted to create a data mesh with the help of dbt Cloud so every team could get their own projects on the dbt Cloud platform within GCP. At that moment, however, they needed support to implement this data mesh structure. That’s where Xebia came in.
dbt Cloud to the Rescue
Xebia’s consultant, Padraic Slattery, brought his knowledge of dbt Cloud to make the data mesh a reality for Kramp. He worked hand in hand with Kramp’s professionals in the Data Office team to set up the infrastructure.
dbt Cloud offered them an integrated development environment (IDE) where analysts could be easily onboarded, removing the need to set up dbt locally. Analysts only had to click a button in their browser to access their project, where they could keep track of their work with the in-built integration in GitHub.
In addition, dbt Cloud offers Kramp data lineage, governance, and documentation. It also connects easily to visualization tools such as Tableau.
To facilitate the onboarding of Kramp’s users to dbt, Xebia’s consultants supported Kramp in defining a dbt Cloud learning journey. In addition, they hosted several trainings that combined Xebia’s dbt-learn course with advanced material tailored to Kramp’s needs. Finally, they also facilitated a dbt hackathon to encourage the development of new solutions.
“Xebia enabled the team to build a solid foundation for our data mesh on the Google Cloud Platform managed with dbt Cloud.”
The Final Impact
Currently, the data mesh is fully running. The platform is fully open for new domains to be onboarded. The process is as simple as reading a Readme file, following a few steps, cloning a Terraform repo, and creating a pull request. Then the Kramp Data Office team approves this request, and the domain gets its new project in 10 minutes.
Once every domain has been granted access to their project, they’re the owners of said project and can:
- Request access to their own source data.
- Grant access to their own data marts.
- Schedule their own data pipelines.
- Schedule their own release to production.
Within the Kramp Data Office team, which oversaw the transformation to the data mesh, the number of active analysts has grown from 0 to 8. On top of that, 30 other professionals have been trained in dbt. This has not only spread data knowledge throughout the company but has also ensured that the data mesh can be maintained to a high standard after Xebia has completed its assignment.
Throughout this project, the Data Office team has increased automation, improved data governance capabilities (data access, data lineage, etc.), and reduced ad-hoc requests. Other departments such as Product, Sales, and Marketing have benefited through the increased ability to own their own metrics, explore granular data, and build customs data models for detailed reporting.
Additionally, teams at Kramp and Xebia liaised to set up a peer-mentoring program and help with the initial pull requests. They also set up shared communication channels to ask questions and share knowledge about the data mesh.