Labellemontagne · Since 2025
A data platform built from scratch
- Microsoft Fabric
- PySpark
- Azure
- CI/CD
- Power BI
Bringing the data of a multi-business group together on Microsoft Fabric, from ingestion to reports.
- 4
- families of sources connected
- 3
- layers: bronze, silver, gold
- 6
- ski resorts covered
Sources
- Ticketing and ski passes
- Hotels and restaurants
- CRM and customer service
- Shops and indoor leisure
Bronze
- Raw data, as received
Silver
- Cleaned and checked data
Gold
- Data ready for analysis
Power BI reports
- Sales for the six resorts
- Customer service
The context. A multi-business group: six ski resorts, hotels, restaurants, indoor leisure venues and shops. Each business runs its own tools. When I joined, in a newly created role, there was no data platform.
The need. Bring this data into one place, so that every team works from the same figures and the group can run its data and BI analyses.
The approach. I started from a blank page, in six steps:
- Choose the stack. Assessing the options, then choosing Microsoft Fabric.
- Map the data. Understanding and documenting the data of each business before writing the first pipeline.
- Connect the sources. Four families: ticketing and ski passes, hotels and restaurants, CRM and customer service, shops and indoor leisure. Many are hosted on legacy servers with no modern connector, and some come from partners.
- Build the pipelines. PySpark notebooks following a medallion architecture: raw data in bronze, cleaned in silver, ready for analysis in gold.
- Make it reliable. Data-quality checks, and a CI/CD process to industrialise deployments.
- Control costs. Automating and managing Azure cloud resources, to make the best use of the platform.
The result. A single platform feeds the group’s reports: sales for the six resorts and customer-service monitoring.