This week, a look at how Shein’s production model has led to brands’ adoption of real-time AI tools, and what that means for fashion in 2024. Scroll down to use Glossy+ Comments, giving the Glossy+ community the opportunity to join discussions around industry topics.
The fast-fashion industry may not be so fast, after all.
This year’s AI boom has improved the availability of real-time data analytics, while generative AI has provided new tools for content and design. At the same time, brands that previously dominated fast fashion, in terms of market share, like H&M and Zara owner Inditex, have been losing out to companies like Shein and Temu that have become known for leveraging lightning-fast, real-time analytics.
On December 13, Zara owner Inditex reported a slowdown in sales growth, with quarterly sales increasing 6.6% to €8.8 billion, or around $9.6 billion. The company’s in-store and online sales hit €32.6 billion, or $34.9 billion, up by 18% versus 2022 and 15% higher than in 2019. On December 15, H&M reported a sales dip of 4% year-over-year, to $6.1 billion, in the three months ending November 30. Annual revenue for the twelve months ending August 31 was $22.3 billion, a 3.6% decline year-over-year.
Sales at 15-year-old Shein, meanwhile, hit $22.7 billion last year, a 44% increase from the $15.7 billion it made in 2021, and 1-year-old Temu is expecting $13 billion in sales this year. With Shein filing for an IPO in November, investors will be able to take a deeper look at the finances and mechanics of the company for the first time — until now, both have largely remained under wraps. Competing companies are increasingly calling out the two companies in earnings reports, saying their strategies are working to gain market share. For its part, H&M is aiming to differentiate by elevating its prices and assortment styles to cater to a more high end customer.
Shein analyzes its customers’ “clicks” across the site to determine what they’re interested in. This, along with social media trends and AI tools, allows the company to predict demand for what customers want using real-time data. If the company gets it wrong by producing an item that doesn’t sell, those items are scrapped and replaced. Temu works in a similar way, across a wider range of products.
“More people will follow that model, obviously not to Shein’s level of extreme,” said Dave Smiley, co-founder of Particl, which provides predictive analytics for companies including Rothy’s, Staud and Skims. “We are slowly moving away from, ‘Going with your gut,’ or, ‘Our president has her ear to the ground,’ and moving toward, ‘What does the data say?’ and, ‘What’s working for the competition?’ It only takes a few failed launches for brands to humble themselves enough to [admit] they do need this data.”
For its part, Swedish fashion company H&M does employ AI algorithms and teams of more than 200 data scientists to predict and analyze trends. Its AI algorithms obtain fashion trend data by capturing information on search engines and blogs. Zara’s use of AI is more similar to Shein’s. The brand’s merchandising teams use data from stores, webpages and the brand app, as well as insights from social media, returns and reviews, to adapt their designs. The entire stock gradually refreshes every four to five weeks. However, neither retailer uses real-time data to consistently create new collections.
AI uses in fashion are becoming more sophisticated
For its part, Particl uses AI to analyze what its clients’ competitors are doing, including when they’re increasing prices, for example. Brands input their own data into the platform, and Particl uses AI to organize and analyze the data. With the men’s shorts brand Chubbies, which had trouble entering the pants category, Particl’s AI and real-time data helped.
“Chubbies is big in swim trunks, and they’ve tried to get into pants in the past,” said Smiley. “So they came to us and said, ‘What’s working with the companies that you’re tracking?’ We’re tracking [brands including] Vuori, Lululemon and Birddogs, so we were able to say, ‘Here are the features that are selling well for them; these are the sizes, the colors and the styles.’ We saw which products could work well for them and took them to market.” The pants launch drove $1.5 million in sales in three days.
Brands’ ability to use AI to access real-time data will allow them to move faster and smarter in the year ahead, in terms of their production. By leveraging both current and past data, they’ll be able to better predict demand in the future. And the explosion of AI models this year shows the potential for more brands to embrace the technology. Lululemon, for example, which has seen results from expanding into new categories like men’s footwear, is using predictive analytics to show whitespaces in new categories.
“Using natural language processing, brands can now simply ask questions, feed in data from their organization, and get the reports and data visualizations they need,” said Kristen Scott, director of data and analytics at digital experience agency Whereoware. ChatGPT, for example, uses natural language processing — through AI, it “understands” text and spoken words. “Even a small mom-and-pop store may use [natural language processing] to forecast inventory, for example, and establish incremental shifts that will drive revenue,” said Scott. “The larger enterprises will just do it at a larger scale.”
Although generative AI is still relatively new, with its boom this year, its capabilities are growing fast. Brands are currently using it to optimize product descriptions, for example, expediting an often tedious process.
AI startup Pre launched its product description optimizer on December 19, which uses AI to both generate and assess the impact of descriptions. Its launch partner, luxury brand Bally, was the first to test the AI tool. “As generative AI becomes more prevalent, brands are finding that it can substantially reduce the cost of content,” Parham Aarabi, founder of Pre. “The trick is to do so in a way that maintains or improves quality.”
“Pre allowed us to automatically generate impactful and engaging product descriptions for thousands of Bally products,” said Nathalie Sisouk, Bally’s chief digital officer. “What would have taken months to do manually was done in just a few days.” According to Aarabi, the tool is reducing the time and cost to generate descriptions — it would have taken 1,000x as long and cost 1,000x as much to do these manually. This is while also increasing the engagement with content by over 50%. Most of the time, as these AI tools use natural language processing, dedicated specialists are not needed to manage them.
When it comes to producing imagery, AI tools are similarly cost-efficient. AI.Fashion, a content creation tool that is in stealth mode and launching early next year, creates photorealistic product imagery based on an image database from brands for brands’ marketing and e-commerce purposes. It is currently working with undisclosed multibillion-dollar brands in the U.S. and Europe, according to the company.
“A lot of the bigger brands on the fast fashion side are utilizing this across their marketing because of the number of styles they’re selling, and this enables them to reduce their costs by up to 60% and increase their conversion by up to 20%,” said Daniel Citron, CEO and co-founder of AI.Fashion.
For the creation of fashion designs, real-time data and AI tools may end up bringing fashion brand-and-customer co-creation to life. Mercury Dasha started in early 2022 as a digital fashion company. In January, it’s launching an AI tool in beta that will allow brands to leverage customers for co-creation opportunites, including choosing colors and seam lengths, for example, before a physical product is made. The AI tool will allow a brand’s designers and the brand’s customers to work together online, while providing an image generator and a design chatbot assistant to help the customer make selections. At the moment, niche brands work better as the product ranges are smaller making for easier real-time adjustments.
“We’re enabling that design process to happen in real-time,” said Mercury Dasha CEO Rita Sheth. “Emerging brands will be able to stop the guesswork about what people actually want.”
Real-time, AI-led data is the future of the fashion industry, with multiple use cases. “If everyone was able to get even 10% better with AI and prediction, we’d see immensely better products launched with consumer-proven demand,” said Smiley. “That means significantly less waste and also less need to discount, because brands would not be launching products that are not wanted.”