A new way to go naked with AI

13/05/2025

At Lush, we’re always looking for new ways to make our customer’s shopping experience better. A big recent change for Lush Lens, our clever little app that scans and recognises our unpackaged “naked” products, is making its way to the mobile tills in our stores.

This case study dives into how this technology works. We’ll explore its impact on scanning naked products, how it’s making transactions in stores faster, boosting our staff’s product know-how, and how we’re managing all the image data using Google Cloud and Vertex AI.

What is Lush Lens? 

We first launched Lens back in 2018 in Harajuku, Japan as our take on digital packaging. The ‘Lens’ feature within the Lush app let you scan and recognise our naked products, giving you all the details you’d normally find on packaging, like ingredients. Meaning we can cut down on excess packaging and waste.

The Naked Product Challenge

Many of our products are handmade and “naked,” which makes identifying them a unique challenge for our technology. 

Before, at the till, our staff would visually identify naked products and manually add them to your basket, either by sight or by remembering a product code (PLU). While most of our team are pros and know every PLU, old and new, this took time, especially when stores were busy. And since we love to bring out new products for you to enjoy, remembering all those new items could be tough.

We needed a better way to capture all the image data. Our in-house team, including Diego Miere from Tech Warriors, created a much faster solution for getting image data, which has really sped up the upload process.

The Solution: Lush Lens x Object Detection on Mobile Tills

To tackle these challenges, we’ve brought Lush Lens to our mobile till system using a technology called object detection. This uses the mobile till’s built-in camera to scan naked products that a staff member holds up, or even straight from the shelf!

We’ve spent years training Lush Lens. What that means is taking hundreds of photos from all angles of a single product, building a comprehensive dataset of our naked products, to train Lush Lens to identify it instantly. AKA object detection.

Once identified, the product is automatically added to your basket on the till. While this is currently only available in the UK and Ireland, we’ll be rolling it out globally later this year.

Impact and Benefits

• Less Packaging: With information available in digital form, we can use less physical material.
• Faster Transactions: Quick scanning means less time at the checkout and happier customers.
• Improved Staff Product Knowledge: Instant identification helps our team recognise products, reduces errors, and improves service.
• Streamlined Training: New staff can learn product identification much faster.
• Reduced Errors: Scanning boosts accuracy in managing waste, billing, and stock.

Data Management with Google Cloud and Vertex AI

We use Google Cloud and Vertex AI to manage and improve our system. Here’s how:

• Google Cloud Storage: This is where we securely store all the scanned image data.
• Vertex AI Platform: This helps us manage and make our object detection models even better.
• Model Deployment: We integrate our trained models right into the mobile till app.
• Continuous Improvement: We can keep training our models using existing product data.
• Data Analysis & Visualisation: We can pull insights from transaction data to make smart business decisions.

By continuously learning and improving our dataset using the machine learning power of Google Cloud and Vertex AI, integrating object detection into our tills really makes the retail experience better for both customers and staff. Fast scanning of naked products speeds up transactions and cuts down on wait times. The system also makes it easier for staff to recognise products and simplifies training.

We manage and own the data generated by our innovative solution, ensuring its continuous improvement and by taking our images of the products in a retail environment.