AI BriefingEnhancing Customer Retail Experience with GPT-4o mini
Key point
Zalando adopted **GPT-4o mini** to strengthen multilingual support and personalized recommendation features, boosting both customer experience and operational efficiency at the same time.
Details
Europe's largest fashion platform Zalando operates the Zalando Assistant, which recommends personalized content to customers and helps them explore products. The existing GPT-3.5-based model had limitations in multilingual support and fine-grained instruction-following, causing results to be too generic when customers requested outfits for specific seasons or events.
To address this, Zalando collaborated with OpenAI and focused on two key areas. First, they built a granular evaluation framework (evals) that allows individual components of the system, such as routing and response generation, to be tested independently. They also strengthened few-shot prompting, providing the model with examples of high-quality and low-quality responses, to improve the model's evaluation accuracy.
They then switched the model to GPT-4o mini, securing multilingual support capability and improving instruction-following performance. This resulted in the following outcomes:
- Product click-through rate increased by 23% and wishlist additions increased by 41%
- 'Not helpful' feedback decreased by 5%
- Service expanded to 25 markets with multilingual support
- Cost-efficient traffic scaling of 12x compared to GPT-3.5
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.