Tyredating is a Michelin subsidiary specializing in web solutions and services dedicated to tires and automotive services. It offers vertical e-retail solutions for garages, auto centers and specialized dealers, aimed at capturing Internet traffic and redirecting it to the physical network in order to materialize in-store sales.
As part of this offer, TyreDating provides a catalog of all existing tires on the market, managed in the Akénéo PIM software.
Issues encountered
TyreDating has identified several critical challenges in managing this catalog
- Significant costs :
Writing and translating descriptions requires significant human and financial resources.
- Incomplete catalog:
Impossible to cover all products, leaving some tire ranges undescribed.
Lack of product filters, even though certain technical information on tires is highly discriminating (winter marking, for example).
- Limited reactivity :
Inability to respond quickly to specific customer requests.
Faced with these constraints, the solution proposed by Algocat aimed to identify relevant filters and systematize the creation of descriptions, while maintaining a very high level of quality.
Methodology in action
The project took place in several key stages, involving both TyreDating's business teams and Algocat's technological tools:
- Gathering business requirements :
- Close collaboration with category managers to understand expectations (text structure to be generated, vocabulary to be used, arguments to be put forward, etc.).
- Definition of technical information to be obtained from manufacturers
- Data structuring :
- Use of scraping techniques to extract essential product attributes for each tire: standards, recommendations, marking, etc.
- Organization of data in a format suitable for automation
- Creating a customized prompt:
- Development of a GenAI prompt reflecting a clear and engaging sales pitch
- Integrate tire specifications into the prompt to avoid hallucinations
- Production and verification :
- Automated description generation using AI
- Team review to ensure relevance and accuracy
A concrete example
Identification of key attributes
Seasons = 4 seasons
Marking = M+S
Standards = EV, Pirelli Noise Cancelling System
Recommended for: electric vehicles, acoustic
Generated text
Results
The benefits of integrating this solution include :
- Improving the quality of descriptions :
- Average score of 4.6/5 for accuracy of information provided
- Average score of 4.5/5 for quality of sales pitch
- Catalog enrichment :
- Additional filters make it easier to find the right tire
- Cost reduction :
- Reduces time spent on manual editing, redirecting it to high value-added tasks
- Increased reactivity :
- Reduced product update and publication times in the catalog
Feedback and learning
Several important lessons have emerged from this project:
- Involvement of the professions:
An important part of the success of a GenAI project is translating the sales pitch and marketing expectations, known to the business teams, into prompt.
- Data quality is paramount:
The quality of the descriptions generated depends directly on the completeness and quality of the technical information retrieved from the manufacturers, so the results are better for premium tires than for budget tires.
- The prompt must evolve:
As sales pitches and marketing expectations evolve over time, a prompt update cycle needs to be put in place.
Next steps
Algocat and TyreDating are planning further developments to manage the personalization of messages delivered according to target, as well as automatic translation.
Conclusion
This use case shows that GenAI is a real step forward for the completeness and quality of online catalogs, provided that business teams are involved and that technical data on products is available to avoid hallucinations.
If you'd like to see this solution in action, please don't hesitate to contact our teams.