At the start of the 2010s, retail merchandising was, at base, a lot of guesswork. The fashion industry made decisions on its seasonal lines once or twice a year, production runs were ordered and goods were shipped. If things went well, and brands guessed right about consumer tastes, that was all well and good. If not, adjustments to the remainder of the stock would be made, and promotions and discounts were thrown out to motivate the consumer.
Retail merchandising was, at best, a hit-or-miss system in its heyday, Nextail CEO and Co-founder Joaquin Villalba told PYMNTS in a recent conversation. However, it is increasingly out of data, and out of step with modern consumer behavior and preferences. Apparel life cycles for products that used to be measured in months are now measured in weeks.
“There is a fundamental shift in mentality, and the necessity for a culture of activity [is that it] can’t just make guesses and hope that they work out six to eight months out,” Villalba said.
What brands used to guide their merchandising processes in the old paradigm requires too much manual input, hinging on data sets and models that are too general, and accruing too slowly — too much like a “fulfilling prophecy,” he noted. Combining artificial intelligence (AI) with data — as Nextail attempts to do with its machine learning-driven retail merchandising platform — does the job better when it comes to better aligning supply and demand.
“If we have data that is coming in closer to the moment of truth (the moment of customer purchase), [then] it seems better to reconsider by earlier decision[s] based on the information I have now,” Villalba said.
Even better, he noted: letting the AI consider the question and automatically optimize for it, instead of making a decision at all.
Adding Granularity To The View
Better merchandising is reliant, first and foremost, on better understanding the consumer. It’s data that retailers already have — what they often lack is a simple ability to put it all together to meaningfully drive a decision. Today, most stores working with limited data, using stock as a hedge against uncertainty, which means retailers push much it everywhere when moving new products.
“We are finding [that] they are shipping out 80 [percent] to 90 percent [of] what they’ve ordered so they don’t miss out [on] any sales in the store. They aren’t sure if their customers will like it (or hopefully really like it), and they don’t want to be caught up short,” he explained.
However, when data is thrown into the mix, what happens next becomes less of an unknown, and it is easier to customize and target where merchandise goes. Instead of shipping 90 percent out, retailers can push out, say, half of their inventory, and see how it performs over the first few weeks — and not just how it does in terms of simple sales, as that information is too general to be useful. Instead of looking at the five or six demographic clusters, which is the norm, properly honed AI can offer a much more granular view across a wide variety of subsets.
“We can see what specific channels are moving, which retail stores, which consumer demographic cluster,” he said. “And at Nextail, if you have 500 stores, then that means they can look [as] represented as 500 separate clusters.”
That level of granularity allows retailers to make much sharper decisions about what to do with the back half of their inventory, based on actual data about where and to whom a product is already selling. It’s not just the product-specific variations on it, though, he noted. Sizing can vary widely, depending on which nation one is in, how many tourists visit a particular sale location or local trend lines. There is a lot of relevant information around which customization is possible, but — without the aid of AI retailers — is unlikely to surface.
Seeing the customer through a sharper lens, Villalba added, means retailers can better tailor and serve their goods to the consumers they are most likely to encounter — the desired front-end experience. Yet, data can also be used to buttress the back end.
Exorcising Phantom Stock
At even the best-run retail establishments, things can go wrong when it comes to managing inventory. Things are mistagged, stock-keeping units (SKUs) get mixed up, phantom inventory haunts the digital archives — falsely reporting items are on shelves when they are not. These all fall under the general heading of stock discrepancies, another area that AI has a better chance at spotting before human eyes.
That is because, he explained, the software is constantly running comparative scans between similar locations, and if a store is diverging from a trend in a big way, the system highlights it. For example, an item is selling at a consistent pace at every retailer but one in an area — this throws up a flag.
“One of the main reasons for that is a deficiency in stock,” he said. “The system is showing theoretical stock, and not products that are actually there. So, the system can both send alerts to actually physically check the stock in-store, and to replenish it if it turns out something is missing.”
Retailers and brands in the apparel business know they need to change, in most cases. When they talk to merchants, they don’t need much convincing that they require greater access to data, and at many more points in the merchandising journey. This is apparel, Villalba noted, and retail brands all up and down the chain — from fast-fashion to premium retail — feel the effects of a faster-moving market, though the details of how that is best managed from a merchandising point of view can vary.
However, the data-driven direction of the market is becoming less arguable. Retailers are increasingly coming into the conversation knowing they need to know more to do better.
“The question we hear now isn’t about whether they want to change. It is about how to identity what levers they can pull, and how to collate their data better so they can actually consider doing things much more quickly and at a much more granular level,” he said.