With 132 stores across France and 1,400 employees, Saint-Maclou faces complex data challenges. From managing sales, installation services, logistics, and inventory, the Lezennes-based retailer must juggle multiple dispersed data sources. A puzzle that significantly hampered its analytical ambitions.
A legacy system that stifles innovation
The starting assessment was unequivocal: Saint-Maclou’s historical decision-support system rested on data flows that were developed and maintained manually. Each new analytics project required precisely identifying the source bases, relevant tables and columns, and then specifically building the feeding pipelines necessary for those analyses.
This approach generated a substantial development and maintenance burden, especially during evolutions. Data engineers spent most of their time on data collection, at the expense of value-added activities like transformation and analysis. License costs piled up, project timelines lengthened, and experimentation with advanced use cases or artificial intelligence became virtually impossible.
“Development load, maintenance load, on-call load, license costs for the tool… Our way of working previously generated significant costs,” explains Salmane Khamlichi, Head of Data at Saint-Maclou.
The move to the cloud and ELT
To break out of this stalemate, Saint-Maclou opted for a progressive overhaul of its data platform, shifting toward a modern cloud-based architecture. The company first deployed Snowflake as the central database, then faced the crucial question of feeding this new infrastructure.
There was no longer room for the traditional ETL (Extract Transform Load) approach. The retailer selected Fivetran for its native compatibility with Snowflake, its proven robustness during a POC, and its ability to handle large SQL Server databases while connecting critical sources.
Today, Fivetran automatically synchronizes data from about a dozen sources to Snowflake, whether hosted on-premises or in the cloud: the Microsoft AX ERP, the logistics tools, the sales and stock management systems, as well as several marketing sources, including the forthcoming Salesforce CRM.
With this ELT approach (Extract Load Transform), the data-collection phase is no longer a bottleneck. Data are loaded raw into Snowflake and then transformed according to business needs.
Measurable gains across the board
The results are spectacular. Gathering new sources, which previously took days or even weeks, now takes only a few hours. Projects stay within a single sprint, avoiding costly back-and-forth that slowed delivery.
On the human resources side, the impact is just as substantial. Data ingestion no longer requires a full-time equivalent. Data engineers can finally focus on high-value operations: transformation via DBT Core and the intelligent exploitation of data. Maintenance issues have become minimal.
The robustness of the solution also provides appreciable peace of mind. Synchronizations are stable and reliable; changes to source schemas do not lead to service interruptions, and teams have access to up-to-date, actionable data with confidence.
Another major benefit: centralization. Where data were previously scattered across multiple business applications, they are now 100% consolidated in Snowflake. This unified view allows for confident planning of future advanced analytical and AI-driven projects. The data team works more closely with business teams (marketing, sales, logistics) and gains in responsiveness.
Finally, the Fivetran platform offers fine-grained control over consumption and precise cost monitoring for data projects. “This enables us to work on sales-forecast topics, rapidly integrate new sources to enrich models, and cheaply experiment with new data science use cases,” summarizes Salmane Khamlichi.
