If haste makes waste, then H&M clearly needs to slow its roll.
The Swedish fast-fashion chain has amassed a staggering $4.3 billion in unsold clothing—up from 7 percent the year before. In an industry where profits are hitched to turnover rates, this volume of surplus doesn’t bode well for H&M’s bottom line. Neither does it improve the burden of a planet already contending with some 26 billion pounds of clothing and textiles entering landfills every year, according to the Secondary Materials and Recycled Textiles Association.
H&M’s saving grace might be the one thing savvier rivals such as Zara have already wielded to great effect: data analytics. The retailer is reportedly using sophisticated algorithms to scrutinize store receipts, returns and loyalty-card histories in a bid to gain insight into what their customers want and, more important, don’t want at each of its 4,288 locations worldwide.
Inventory optimization, powered by numbers rather than instinct or experience, is obviously the goal here. As a rule of thumb, about 80 percent of sales come from 20 percent of a company’s assortment, said Karen Moon, CEO and co-founder of Trendalytics, a fashion-centric data-analytics firm based in New York City.“Although the numbers might vary for different people, it really is true within the commerce landscape where your bestsellers are driving a disproportionate amount of your revenue,” Moon said. “And so getting that right is really important.”
Trendalytics looks at various data streams, including social-media buzz and online searches, and breaks them down by product type so it can pit, say, denim jackets against bomber jackets or trench coats to gauge demand and supply. “What you can do is predict where consumer demand is headed and make better assortment opportunities as a result,” Moon said.
The right kind of data can help brands and retailers optimize their assortments, hedge fewer inventory “bets” and squeeze the same returns from a narrower but more targeted product offering, Moon said. Understanding consumption patterns can allow companies to time their product drops with maximum customer interest. Should swimsuits grace racks in March or June, for example? When are people in the market for rain boots? When should markdowns begin for sundresses?
Plus, fewer unsold clothes means less waste—a win for the environment. “Everything around excess has to do with uncertainty of demand,” Moon said.
Target, for instance, is testing a proprietary app that serves as a direct line into the minds of a select pool of customers. Through Studio Connect, designers can glean feedback about specific products or crowdsource ideas for future ones. Eventually, with an expanded user base, Target might use the data from the app to curate the offerings of individual outlets.
Zara uses radio-frequency identification (RFID) to tag its products and manage inventory flow. Because the retailer can quickly determine if a certain style is flying off the shelves or languishing on the sales floor, it can rejigger its assortments accordingly.
Indeed, for companies that are ramping up their sustainability goals, reducing overstock is one low-hanging fruit Big Data can easily tackle with relative ease, said Jessica Graves, founder of Sefleuria, a boutique firm that employs machine-learning strategies to help fashion businesses scale. “It also takes away some of that pressure that we put on the supply chain to get more stuff produced in less time for less money,” she added.
Getting a better handle of the kinds of garments customers want (and how much of them) might even relax some of the tensions between the demand for ever-cheaper clothes and the welfare of the people who make them.
“If you reduce some of that pressure of, ‘We need to get x many garments,’ then your factories aren’t incentivized to sub-contract to factories that you’ve never been to, you’ve never heard of or you might never have approved,” Graves said, referring to the small, often unregulated “shadow” facilities that populate the gray economies of producer countries such as Bangladesh.
And raw numbers aren’t the only way of predicting what’s in, out or on the cusp of a comeback. IBM and the Fashion Institute of Technology are training a visual search engine, known as Cognitive Prints, to filter results by design element, pattern, color, genre or era. (Some possible examples: “1920s white wingtip Oxford shoes” or “1980s shoulder-pad prom dresses.”)
Similarly. a French startup called Tagwalk aspires to be the “Google of fashion” while parlaying popular keyword searches into the type of consumer insight brands crave. Sotheby’s recently acquired a deep-learning company that specializes in image recognition and recommendation to boost its auction searches. Even Asos has a Style Match app search that allows users to search the e-tailer’s inventory by uploading a photo and getting similar looks in return.
It’s the rare visionary, Graves said, who looks at data “not just as numbers and counting and business intelligence but creating a product experience and becoming a user tool.”
Recommendation systems are only going to get increasingly sophisticated, especially when you throw artificial intelligence (AI) into the mix. AI could even help shift from suggesting products to designing them. Researchers from Adobe Research and University of California San Diego, for instance, are experimenting with neural networks that could someday create bespoke garments based on a buyer’s personal taste.
This could pave the way for a “new type of recommendation approach that can be used for recommendation, production and design,” said Julian McAuley, a professor of computer science and engineering at the school. “These frameworks can lead to richer forms of recommendation, where content recommendation and content generation are more closely linked.”
For companies leaning into machine learning, however, Graves warns that data cannot completely supplant human intuition and the good ol’ gut check. A person can buy only so many cold-shoulder tops, after all.
“Most people assume using lots of data and smart algorithms will improve predictions about what inventory customers will buy, when algorithms are often mirroring the past forward,” she said. “If consumer behavior changes or the market changes, the predictions may end up less relevant and less accurate. How the algorithms allow for serendipity and assumes the customer changes over time will be key.”
Personalization and waste reduction aside, data can also help shine a light on some of the more opaque portions of the garment industry, where social and environmental abuses can flourish unchecked. Provenance, a London-based software service uses blockchain, along with “smart” labels that interact with mobile devices using near field communication, to help brands trace the origins of their products and, in doing so, promote greater transparency in their supply chains.
Blockchain breeds trust because its ledger technology is decentralized, meaning there is no one owner of the information contained within a “chain.” Neither can these blocks of information be edited or deleted without leaving behind a “scar” as a signal that something has been altered.
Jessi Baker, the founder of Provenance describes blockchain as a “real-time Facebook timeline” for a product as it’s moved from one point in the supply chain to another. “It’s tracking, like FedEx for food,” she added.
For Provenance, transparency will one day be a “fundamental requirement” for retail success. Brands will need to verify the sustainable and ethical claims they make about their business, said Neliana Fuenmayor, founder of A Transparent Company, a tech agency working with Provenance on several pilot projects.
“In the future, we would like people to make more conscious decisions about what they are buying and consuming not just based on what their stomach is telling them or what they can afford but based on the story behind the item and how it was created—that is a really powerful way to buy,” Fuenmayor said. “This means brands and global businesses will likely fail if their customers cannot easily access product information.”
Moon agrees. “Transparency is going to be really important in the future,” she said. “It’ll be an expectation almost.”