The challenge
Sales and inventory data was spread across the ERP, the in-store point-of-sale systems, the online store and dozens of spreadsheets that each department maintained on its own.
Every week a team consolidated the data by hand to prepare reports. It took days, depended on a few people and the figures did not always match between departments.
The same product appeared with different codes and descriptions in each channel, so nobody could say for sure what was selling and what was missing in each store.
The solution
We built ETL pipelines on Microsoft Azure that automatically extract data from every source, transform it and load it into a unified data model for sales, inventory, products and stores.
We added data-quality rules (duplicates, out-of-range values, dates and currencies) and alerts when a source does not arrive or arrives incomplete.
Language models standardise the catalogue: the AI suggests which master product each description belongs to and the team validates uncertain cases, so the catalogue is cleaned once and stays clean with every load.
Power BI dashboards were published on top of the unified model, which is also the foundation for an AI agent that answers business questions in natural language.
Results
- Sales and inventory reports refreshed every day, with no manual consolidation.
- One version of the numbers for sales, operations and finance.
- A product catalogue standardised across channels and stores.
- Data ready for advanced analytics and AI agents.