Algorithms already tell us what to watch, now they’re telling us what to wear.
It’s not that consumers don’t appreciate the help putting fashion outfits together (style influencers still lord over social media), but online retailers don’t have the bandwidth to exponentially scale up outfitting suggestions like artificial intelligence does.
Meet the man behind “Shop the Look.” Rohan Deuskar is the founder and chief executive officer of AI-powered digital merchandising system Stylitics—an AI-powered digital merchandising and styling technology, now used by 100 million consumers shopping fashion and home furnishings at Puma, Macy’s, Walmart, Kohl’s, and Revolve among others. The visual outfitting company recently closed an $80 million Series C fundraising round, bringing its total funding to $100 million, and it recently released its fourth-generation version with advanced algorithms and expanded capabilities that include advanced personalization features for better intent and behavioral mapping, with 10 times more speed across content creation systems.
The platform recommends outfits and bundles in over 50 billion shopper sessions a year, resulting in a 23 percent increase in units per transaction and a 21 percent increase in average order value for its partner brands and retailers.
Sourcing Journal checked in with Deuskar, who explains how his original model, which pivoted from B2C to B2B, is coming full circle with a hybrid personalization model for retailers. And how Stylitics‘ outfit suggestions keep consumers from getting overwhelmed by the Paradox of Choice.
Related Stories
Sourcing Journal: Going back to the beginning, how did your background spark your idea? Were you a fashion guy or tech entrepreneur guy?
Rohan Deuskar: I had worked at another high-growth startup in Chicago in the mobile messaging and tech space, but in business school I interned at Amazon where I saw retail, data and tech all coming together.
I realized that while I wasn’t the most stylish guy in the world, I was buying clothes, but I kept buying the exact kinds of things I already owned. I thought: ‘I wish my closet had a brain’ to tell me what I had, what I needed, and what to wear and when. My co-founders and I ultimately created the original version of Stylitics, a pioneering digital closet platform where you could mix and match new products with what you already owned, build outfits and get a style profile. The app (later rebranded as Closet Space) was very successful around the world, but it wasn’t shoppable.
SJ: So you brought the concept to retail?
RD: Ten years ago, brands were feeling like they didn’t have enough data, unlike today where we often have too much and don’t know what to do with it. So we created a sort of Nielsen for fashion on the back end telling brands what consumers had in their closets, with the smart closet platform on the front end. Retailers, meanwhile, were trying to solve for how to show consumers outfit recommendations, so we went back to our original investors and created the first visual outfitting solution on the market.
SJ: How does Stylitics translate its data to specific brands that must keep their brand DNA and consumer preferences in mind?
RD: Our system has 100 billion sessions worth of data—the largest database of the pairings that we’ve seen. We have massive shopper data, and we have our AI and algorithms that can produce really good full outfits. Anyone can build a dumb AI that just puts stuff together or makes add-on recommendations. But shoppers don’t want to see more product in boxes. We make sure all of our outfits reflect the brand’s specific vision and the merchant-specific considerations.
SJ: Algorithms are taught, so how can brands help set those merchandising recommendations?
RD: We already have enough experience from thousands of brands to know what does and doesn’t go together, but it also has to make sure that when the merchants say, ‘We really care about clearing through this type of inventory, or we really care about our private brands, or we have a new designer who we don’t to pair with the older stuff—except if it’s maternity and basic colors, in which case you can.’
The real challenge is making it extremely high quality so there is trust. A consumer doesn’t know what’s happening in the background, they just think Nike or Bloomingdale’s is suggesting outfits. It’s extremely hard and we’re proud of having built a system that can take all of those merchandising considerations and put it into the technology.
We do have humans working with their humans!
SJ: And how is AI augmenting that human knowledge?
RD: We’re helping them scale their vision. What we are replacing for retailers is what they might have done in studio for, say, 15 visuals. Retailers would spend $200,000 doing a bunch of lay-down photoshoots those would be the outfitting recommendations served back in the day. [With Stylitics], they’ll have 200,000 outfit recommendations on site. Retailers were never going to spend $200 million in studio doing that.
SJ: Are you relaying data back to your stores to make information less static?
RD: This is a growth opportunity for us. Our customers have something like 40,000 stores, collectively, and maybe 200,000 associates at the low end. But if associates get a one-time guidance of what to put on the mannequin, those products might sell out, or go off trend, or be different in various markets. So, we’re starting to experiment on behalf of our customers with sending associates a quick readout of the top-performing outfits in their area for certain products. A lot of times associates can’t really guide you, so we’re looking at ways to enable that associate to be able to help you without having that deep expertise themselves. And also for customers to self-service.
SJ: How is Stylitics being personalized for retailers to accommodate shoppers with different purchasing patterns?
RD: Most of our customers have moved onto our personalization tech and we’re rolling more out. For example, a jacket will have 50 outfits or 100 outfits behind the scenes, but there’ll be a huge variety—some super casual, high price, low price, minimalist to super glam with everything in between, and outfits for lots of different occasions.
We want to make sure that when you come to a site—based on what you’ve been browsing, or outfits you’ve engaged with or responded to and some of the other patterns we’re detecting—you’re going to see outfits that will be different than for another shopper. Then we’re factoring in the best-performing outfits or best-performing for the kind of weather we suspect you’ve had because of broad level geolocation.
It’s also a way to push certain items that are not selling, say, ruffles, before you have to put them on clearance. We won’t show it to the minimalist person who’s not going to appreciate it, but we will make sure to show it to the person through the personalization aspect who is best suited for it. So the merchant who’s bought $200,000 or 200,000 units of ruffles, gets more traffic to the product page, getting in front of the right shopper, and it’s totally free. It’s a new superpower for these merchants and we’ve only scratched the surface.
SJ: Can the system make recommendations that mix and match with items a shopper already purchased? Sort of like your original smart closet idea?
RD: For some retailers, there is the idea of showing you outfits for things you’ve previously bought. If we do know what you browse and what you purchased on the product detail pages, we’re boosting certain things based on your past purchases.