2024 DataIQ Awards – Data-Enabling Solution of the Year (Brand): Burberry

Burberry’s Visual Product Platform has been awarded Data-Enabling Solution of the Year (Brand) for its ability to significantly improve efficiency and scalability.

Burberry’s Visual Product Platform (VPP) innovates product-centred processes by leveraging AI to automate image analytics and captioning, significantly impacting efficiency and scalability. Processing tens of thousands of product images annually across all international markets, the VPP has saved more than £500,000 by reducing manual work hours annually. By fine-tuning AI models on 500,000 images and 10,000 captions, it offers visual similarity search, custom tagging, and AI-generated captions, enhancing product discovery and SEO. This strategic innovation streamlines operations and accelerates market readiness and sets new industry standards, showcasing Burberry’s commitment to combining technology with fashion.  

 

Merging fashion and tech 

Burberry has harnessed the power of data and AI to innovate the product launch process, transforming the way the team works with visual and textual information. The newly launched VPP has enabled streamlining of operations, unlock new use cases, and drive efficiencies that would have been impossible to achieve manually.  

Annually, Burberry launches thousands of products across 50 countries, contributing to £3 billion in sales. The journey of each product from concept to consumer is complex and visually driven, starting with sketches and 3D models and culminating in the product photos shown in the online store. As these products advance towards their market debut, the need for detailed textual information escalates. Product descriptions, image alt text, and precise image tagging are crucial for accessibility, SEO, and facilitating effortless product discovery. Traditionally, accomplishing these tasks was labour-intensive, demanding over 2,000 hours of manual work each year (a process both time-consuming and subject to constraints), making it impractical to write accurate image descriptions and captured detailed product tags for every photo on Burberry.com.  

Identifying this bottleneck, Burberry’s data science team developed the VPP. By integrating modern AI, such as image embeddings, object detection, and multimodal LLMs, and fine-tuning these on an extensive internal dataset of over 500,000 images and more than 10,000 manually curated captions, the team has crafted a comprehensive solution addressing all product-related image tasks. The VPP offers services through APIs and a custom front-end including visual similarity search, custom image tagging, and LLM-powered image captioning, thereby becoming a cornerstone of the Burberry data science operations, and powering essential internal tools for product recommendations, merchandising planning, and .COM copywriting.  

Products based on VPP have been deployed in production, including Burberry.com and email marketing, enhancing millions of online journeys every week. Through A/B testing, the Burberry data team has demonstrated that internal solutions consistently and significantly outperformed commercial third-party offerings in personalisation, searchandising, and image analytics. This led to measurably higher conversion rates through a more personalised user experience and improved accessibility, generating incremental revenue.  

The journey to develop and implement the VPP faced significant challenges, especially in technical and legal areas. Deploying generative AI models, such as multimodal LLMs for image descriptions, in a production environment on an international scale demanded a collaborative effort across engineering, legal, copy, and DEI teams. This cooperation was vital in navigating the complexities of legal frameworks and ensuring compliance with internal guidelines and international standards to guarantee the generated consumer-facing content is relevant, inclusive, and brand-safe.  

The impact of the VPP on Burberry’s operations has been profound. By automating tasks that previously required extensive manual labour, Burberry has unlocked efficiencies and enabled new use cases that were once deemed unfeasible. This transformative approach has not only saved over £500,000 in manual work but has also significantly accelerated the product launch process, improved the accuracy of Burberry’s product tagging, and enhanced the overall customer experience on Burberry.com.  

The VPP is more than just a technological solution; it is a strategic asset that underscores Burberry’s dedication to innovation, efficiency, and customer-centricity. It demonstrates how creative applications of data and technology can solve complex business challenges, drive value, and set new industry standards.  

Membership

Interested in
joining DataIQ?

Join our membership network 
of over 250 global senior data 
AI leaders.

Join our membership network of over 250
global senior data and AI leaders.