
H&M is losing customers—can Big Data bring them back into the fold?
The Swedish apparel retailer certainly has had a hard time of it lately. Sweatshirt controversies and delivery problems aside, foot traffic has nosedived, sending H&M’s operating profit careening 62 percent in the three months through February and yoking the fast-fashion chain with $4.3 billion in unsold clothing. Shares have slipped a precipitous 56 percent over the past three years.
But artificial intelligence (AI) could help H&M engineer a reversal of fortune, the Wall Street Journal reported Tuesday. To work its way out of this slump, the retailer is said to be employing algorithms to scrutinize store receipts, returns and loyalty-card histories so it can better customize its inventory at individual locations: a bigger selection of tank tops and skinny jeans in one store, perhaps, or more cashmere sweaters and pencil skirts in another.
It’s a strategy that requires far greater forethought than traditional stocking practices, which hitherto have relied on a one-size-fits-all philosophy or a store manager’s instincts. Drilling down to such a granular scale might also prove difficult given both H&M’s size and stature. The retailer’s fleet includes 4,288 stores across the globe, or more than double Zara’s 2,127.
Still, H&M is not the first brand to use data to court its clientele. Zara is a pioneer of using data analytics to track demand and supply on a localized, real-time basis. Gap mines past purchases to suggest complementary styles that its customers might like. Uniqlo is tapping Big Data to figure out with needle-sharp precision what its customers want—more packable puffers? Less gingham?—in each of its stores, curtailing excess inventory. Even virtual assistants like Amazon’s Echo Look are employing machine learning to make style recommendations and trigger acquisitions.
AI could even predict—or even create—the next fashion trends. IBM and the Fashion Institute of Technology are training a visual search engine, known as Cognitive Prints, to hone in on images with specific elements such as a surplice collar, a Breton stripe or a cold-shoulder sleeve. Users can filter results by designer, era or genre (“1970s Halston formalwear” might be one example.)
Cognitive Prints can also design patterns based on visual inputs: the sharp curves of the Eiffel Tower, say, or a panorama of the night sky. Eventually it might even dream up whole garments based on keywords like “white,” “jumpsuit” and “lace.”
“As AI progress continues to advance, fashion [will] see more transformations,” Priyanka Agrawal, an IBM Research India scientist who worked on the project, told Smithsonian Magazine. “For example, with the rise of conversational agents and virtual reality/augmented reality technology, it should not be long until users can not only query fashion catalogs but also interact, iterate and [be inspired by] the technology.”
Meanwhile, researchers from Adobe Research and University of California San Diego are tinkering with neural networks that they claim can one day create custom apparel based on a buyer’s preference.
If successful, such a model would offer 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.”
In other words, it’s Big Data’s world; we’re just living in it.