Nous contacter

Matalan

Une avancée majeure de l’IA dans le retail

Le modèle de GenAI qui a transformé l’expérience d’achat en ligne d’un retailer.

A person holding a smartphone, viewing a product page for Grey Stripe Thinsulate Gloves. The screen shows the product description, price, and purchasing options. In the background, a laptop and a small potted cactus are visible, suggesting a casual or work-from-home environment.

Secteur

  • Retail

Offre

  • Croissance par la data et l'IA

Technologie

  • Google

Présentation de Matalan

Matalan

Matalan est un retailer omnicanal basé au Royaume-Uni, spécialisé dans le prêt-à-porter féminin et masculin, ainsi que dans l’équipement de la maison et le mobilier. L’enseigne exploite plus de 280 magasins en propre et en franchise au Royaume-Uni, en Europe et au Moyen-Orient, tout en développant rapidement son canal de vente en ligne.

Une production de contenu à grande échelle

Matalan

With such a wide range of products, the company’s copywriting team was tasked with creating a large volume of product descriptions on a continual basis. Matalan wanted to explore whether generative AI technology could alleviate some of the workload, deliver better search results to its customers and ultimately lead to an uplift in revenue. 

  • A tablet screen displaying the Matalan website homepage, featuring a 'Bring on the Bank Holiday' event with up to 30% off. Below the main banner, various promotional offers for different product categories are shown, including women's, men's, home & garden, and kids' items. The design is bright and inviting, with a focus on seasonal sales and department browsing.
  • A display inside a retail store featuring the 'New Papaya Style Edit' collection. Mannequins are dressed in neutral and green tones, showcasing casual outfits with accessories like a green crossbody bag. Racks of matching clothing surround the display, with a modern, stylish vibe.
  • A tablet screen displaying the Matalan website product page for a multicolored ditsy viscose mini tea dress. The page shows multiple images of a model wearing the floral dress, with details on pricing, size options, and delivery methods. The screen is in landscape mode, providing a clear view of the shopping interface.

GenAI, supported by humans

Matalan

In just three weeks, we built a generative AI and large language model (LLM) solution with Google Cloud Platform's Gemini and PaLM 2 on Vertex AI.

The bespoke tool is capable of creating product descriptions using product metadata for Matalan's full product range. Configured to read product imagery, it selects details from images to include in the descriptions. To ensure accuracy and quality control, the copywriting team oversees the automated descriptions.

As the UK’s first retailer to be leveraging generative AI in this way, this is a hugely exciting moment for Matalan. Not only does it demonstrate innovation into new and so far unexplored areas, but it shows our commitment to putting our customers and the shopping experience first — whether they’re shopping on our website or through the Matalan app.

— Ali Jones, Chief Customer and Omnichannel Officer, Matalan

Before and after comparison of a product description page for Grey Stripe Thinsulate Gloves. The 'Before' version shows a simpler layout with minimal details, while the 'After' version provides a more detailed description, including product features, ways to wear, and additional information for a richer shopping experience.

Results

Matalan

The GenAI model is the first of its kind for the UK retail sector and has already provided several benefits to Matalan:

  • It creates 100 descriptions in 30 minutes (versus 100 per day by the copywriting team).

  • It generates rich, accurate product descriptions with minimal data input, leading to improved SEO and conversion rates.

  • It increases copywriters’ efficiency, freeing them up for more strategic work.

  • It enhances the customer experience with longer and more emotive product descriptions.

Nous contacter

The experience innovation company