Before data can drive every aisle, warehouse and recommendation, someone has to build the foundations

Intermarché, France’s third-largest food retail network, has long known that the retail battle is won in the details: the right product, in the right place, at the right time. But running thousands of stores, anticipating stockouts, personalizing recommendations and accelerating e-commerce within the same strategic sequence takes far more than Excel spreadsheets and gut feeling from the field. It takes data: structured, governed, put to work.

To deliver on this ambition, Intermarché creates a Data Factory: a unit dedicated to industrializing the value extracted from data, able to run several initiatives in parallel. But a Data Factory without a reference architecture is a construction site where every team invents its own foundations. Technical debt builds up silently, and what was meant to speed things up ends up slowing them down.

Nobori Partners joins the program as its first data architect, to get things back on course and enable the real-time use cases.

The work starts upstream of the code, in business workshops. For each initiative, the task is to understand the business process involved, identify the available data, assess its maturity and choose the right patterns: how to ingest, how to transform, how to deliver. This initial scoping avoids bad surprises during development and aligns the data and business teams before the first pipeline is written.

On the ML and AI front, three structural use cases are covered: a product recommendation engine, a tool comparing performance across stores, and customer insight models. For each one, the architecture covers the feature pipelines, the training pipelines on Databricks and Spark, and the delivery interfaces to the consuming applications through Azure Functions.

BI reporting is the second major domain. Supply chain teams need to see stock levels, stockout risks and logistics flows in real time. Store managers want their own dashboards: clean, consistent, actionable. Nobori Partners scopes the reporting needs with the business teams and defines the architecture of the data flows feeding the Power BI dashboards.

The third domain is real-time streaming for e-commerce. Here, latency matters: an online order, an updated stock level, a recommendation displayed within seconds. The architecture relies on Kafka, Solace and Boomi iPaaS to orchestrate event streams and make sure data moves at the speed customers expect.

Across all of this, Nobori Partners structures the data architect function itself within the Data Factory: architecture review templates, a pattern library by type of project, a standardized onboarding process. When a new use case comes in, the team no longer improvises; it follows a proven path.

Over the years, more than fifteen initiatives have gone from idea to production within this framework. Recommendations, logistics management, customer insight, e-commerce streaming: each one benefits from a well-designed architecture, documented governance and a consistent technical foundation that makes evolution and maintenance easier.