Leroy Merlin: Using a Knowledge Graph to Improve Product Recommendations

Adeo, the parent company of Leroy Merlin, embarked on a transformative journey with us to enhance their product catalog. This case study explores the challenges they faced, the innovative solutions we provided, and the remarkable results achieved.

The Challenge

Adeo's dedicated team, consisting of taxonomists and category managers, was committed to meticulously curating their product catalog. Utilizing Poolparty, an advanced ontology management tool, their objective was to develop an extensive knowledge graph.

This graph was instrumental in listing a diverse range of classes and properties, setting the stage for a comprehensive e-commerce product recommendation system, akin to an in-store advisory experience.

The Vision

The heart of this project was the creation of a sophisticated knowledge graph. By interlinking products through a network of relationships, we aimed to build bridges across product categories. This not only enhanced product discoverability but also elevated the customer experience by offering personalized and contextually relevant recommendations.

Bottlenecks Encountered

Despite their relentless efforts, Adeo's team confronted a significant hurdle. The manual process of identifying new classes and properties had reached its maximum, with hundreds of hours invested and diminishing returns.

Lettria’s Solution

Our approach revolutionized the enrichment process. By harnessing unstructured content from various sources – including Leroy Merlin's blog, product sheets, YouTube channel subtitles, and industry-specific blogs – we fed this rich data into the existing ontology. The Lettria platform played a pivotal role, analyzing thousands of documents to suggest new ontology elements.

This automatic ontology enrichment, executed seamlessly through the Lettria platform, aligned free text content with the initial ontology, unveiling a plethora of enrichment opportunities such as new classes, relationships, and attributes.

Results and Impact

Our solution yielded tens of thousands of enrichment suggestions, derived from diverse methodologies including n-gram analysis, leveraging Lettria's knowledge base for contextual suggestions, LLM capabilities for contextual insights and synonym identification, and a specialized machine learning model for abbreviation detection.

The impact was profound. Adeo's taxonomists and category managers were equipped with an intuitive interface to review and approve these suggestions. In several instances, over 60% of our proposals were integrated, allowing Adeo to augment their product ontology rapidly.

Within a matter of hours, thousands of new classes and properties were added, significantly enhancing their e-commerce platform's recommendation capabilities and overall customer experience.