In 2021 Amazon passed Walmart as the top apparel retailer in the United States, thanks in part to a pandemic-linked boom in online ordering. Today, the global fashion ecommerce market is worth in excess of US$800bn, with Statista predicting it could be worth over US$1.2tn by 2027.
The rise of AI has had a transformative impact on the way the world works, and now – thanks to advanced AI models – the technology is poised to revolutionise the way we shop. According to a blog by Apoorv Chaudhri, Director of Computer Vision and Machine Learning at Amazon Fashion, valuable insights unlocked by AI have been helping the company improve the customer experience and address the size and fit challenges that are part of shopping for fashion online.
Here, we look at how Amazon is transforming the fashion industry, using powerful AI technology to provide better experiences for its customers.
How Amazon’s AI-driven size recommendations create greater personalisation for consumers
Amazon is using AI and machine learning models that enable it to provide better recommendations on product detail pages.
Central to this is a deep learning-based algorithm to help each customer find their best-fitting size in any style. According to Amazon, this algorithm considers the sizing relationships between brands and their size systems, a product’s reviews and other details, and a customer’s own fit preferences to recommends the best-fitting size for a customer.
The algorithm anonymously clusters together customers with similar size and fit preferences, and products with a similar fit. From there, the algorithm learns from millions of product details, such as style, size chart, and customer reviews and billions of anonymised customer purchases. It also takes into account sizes bought and kept by similar customers for the same product, or for similar-fitting products.
“We’ve learned that customers are more likely to purchase and keep an item when a size is recommended for them,” Chaudhri says. “The size recommendation system analyses millions of data points every day and generates billions of size recommendations each month for hundreds of millions of customers across 19 locales around the world.”
Fit Review Highlights offer relevant customer feedback
Another feature that helps Amazon customers find the right fit is AI-generated Fit Review Highlights. Amazon creates a review highlight for each customer based on their recommended size using common themes across reviews, with the goal to make it easier for customers to get personalized size guidance.
The feature tells a customer whether to size up or down in a particular style based on reviews from customers who have purchased the item in the same size.
“We use the latest and most advanced forms of AI, like large language models (LLMs), to extract details from customer reviews, such as size accuracy, garment fit on specific body areas, and fabric stretch,” explains Chaudhri. “We then use AI to summarise these details in an easy-to-read review highlight. The highlight guides each customer to the most relevant information, so they don’t have to manually parse through hundreds of reviews for each item.”
Reimagining clothing size charts with trusted AI-powered data
Amazon is also using AI to make size charts more accurate and useful for customers, while showing customers relevant information in a more visually engaging way.
By leveraging LLMs, Amazon automatically extracts and cleans product size chart data from multiple sources. This data is then transformed into standardized sizes, with duplicate information being removed and missing or incorrect measurements being auto-filled, resulting in a more accurate and consistent size chart.
AI Fit Insights Tool enables brands to improve their product offerings
As Chaudhri describes, brands and selling partners can benefit from all of this innovation, too. “Understanding why customers returned an item can be a mystery, but with our new Fit Insights Tool, Amazon can do the heavy lifting for brands and offer insights into the fit of various products,” he adds.
Amazon’s Fit Insights Tool uses an large language model to extract and aggregate customer feedback on fit, style and fabric. It contextualizes returns and size chart analyses with customer reviews, using machine learning to identify defects in size charts. By leveraging this data, brands can better understand customer fit issues, improve how they communicate sizing to customers, and even incorporate the feedback into future designs and manufacturing. This helps brands reduce fit-related returns and more accurately list their items for customers.