Artificial intelligence has come to mean more to retailers than just chatbots and recommendation engines. AI has found its place as a critical tool, particularly for companies managing a fleet of stores.
That’s certainly the case for Puma and PVH Corp., which have both come to rely on the technology in the critical months of the coronavirus pandemic.
Katie Darling, Puma’s vice president of merchandising, and Kate Nadolny, senior vice president of business strategy and innovation at PVH, weighed in on how their companies used AI to improve forecasting and merchandising during a tough year that has vexed brick-and-mortar retail like no other.
Nadolny explained to host Prashant Agrawal, of Impact Analytics, that when PVH Corp. began its journey with AI forecasting a couple of years ago, the company wasn’t entirely sure about what it would entail.
“We identified the clear need and opportunity for us to be smarter about how we’re making our forecasting and prediction decisions,” she said. “But we weren’t really sure about what the tools and capabilities were that we needed.”
As the parent company of Van Heusen, Tommy Hilfiger, Calvin Klein, Izod, Geoffrey Beene and more, PVH had more to weigh than a chain of single-branded stores. It was dealing with a multibrand portfolio, with different target customers and locations, bringing an extra layer of complexity. But it’s the sort of challenge that AI meets head on.
Saint Laurent RTW Spring 2021
The company first looked to brainstorm with tech partners over areas like assortment and allocation, but COVID-19 quickly changed the priorities.
Suddenly, Nadolny said, PVH realized that it needed to make swift decisions, as its stores contended with varying rules that meant some stores closed, while others remained opened or swung between the two, as infection rates changed. Meanwhile, the rules of retail were being rewritten, as consumer behaviors morphed.
This was the period when stores were exploring curbside pickup and retailers that never before offered appointment shopping suddenly raced to meet new customer expectations. And such services may not work equally well in all areas, particularly in regions hard hit by the economic downturn, or perhaps work best for certain product categories or customer segments, which can vary by store.
Nadolny explained that knowing where and how to shift inventories or change pricing in real time, AI is simply faster at crunching the data and pulling insights than humans.
Darling agreed. She discovered at Puma that granular planning, down to the store level, across multiple doors is impossible without artificial intelligence. Additionally, “it can find patterns you’re not looking for,” she explained, especially when compared to the way people dig through spreadsheets.
It’s not only slower, but also less efficient at spotting identifying critical insights.
If one product sells out, what’s the next best item in stock that can fill the gap? If a certain item performs well, but what’s really leading sales are the smaller sizes in that particular style or stockkeeping unit, could a human staffer pinpoint that? In a normal year, such questions would point to missed opportunities, but in 2020, those insights can determine survival.
“The idea of using artificial intelligence to help us make smarter decisions, whether it be at a category level, a collection level or even down to a size level is really important,” said Darling.
But before AI can be really useful in the retail setting — or in any setting — the fundamentals need to be in place. Nadolny pointed out that AI initiatives need to start out with good data, which was one of the PVH’s biggest early challenges.
“As the saying goes, ‘Garbage in, garbage out,’” she said. “So how do you make sure that your data is right? That attributes are right, that the information that we have is correct and aligned? So while we have quite a bit of data that’s very, very useful for us, it’s not always in the same place, in the same structure, in the same format.
“As we start to move towards being able to better utilize these types of tools, internally we’re spending a lot of time focused on the clarity of our data governance and data structure, so we can therefore take that information and utilize that appropriately in the tool,” she added.
The human element also remains important, Nadolny noted, in that staff should have appropriate training on how to best use the tools for the business.
The process could be a challenge for the humans, Nadolny acknowledged. It can even feel like a loss of control, but ideally they’ll come to see and appreciate the tech. “The machine can really learn more quickly and adapt to what’s happening in the space — more so than we can in our Excel-based toolset that we have today,” she said.
Source: Read Full Article