Carrefour, a global retail giant, collects valuable opinions from its customers in Spain, facing the challenge of analyzing a vast array of varied comments. The critical need to process this data to optimize the customer experience and prevent inconveniences in their stores motivated the search for an innovative solution in collaboration with WhiteBox, within the SOFIA program of the University of Navarra.
The strategy began with a close collaboration with Carrefour to establish a detailed ontology of comments, categorizing aspects such as product quality and price, comfort, and specific areas of the store. Through this ontology, we developed artificial intelligence models, using open source tools such as scikit-learn and Hugging Face Transformers, to classify comments, evaluate emotional tone and detect references to competitors.
The project was encapsulated with MLFlow, ensuring efficient management, and was deployed in production using Docker containers. A dashboard in Looker Studio was implemented to allow the Carrefour team to explore the processed data, facilitating the extraction of valuable insights and the visualization of trends and patterns.
Through this project, Carrefour revolutionized its feedback analysis method, moving from a limited manual focus to an advanced automatic system. This system not only alerts about potential problems in real time but it also reveals hidden opportunities in customer feedback, inspiring new products and services.
- More than 40,000 comments analyzed monthly by our models.
- Savings of more than 20,000 hours of manual work per month in time spent reading and processing comments.
AI-Driven B2B Lead Prioritization
Automated Enterprise Document Field Extraction
