After first launching in July 2018, AI-powered clothing company Choosy quickly decided to pull back its operations and rebrand the company before relaunching last November. Now, a year in, the company is on track to total $6 million in sales for the year, thanks to a revamped strategy around inventory and production.
At launch, Choosy was trying to be more of a platform, selling lookalike pieces of popular styles — like the spring 2018 Jacquemus dress worn by Kim Kardashian — at a fraction of the originals’ cost. It didn’t hold any inventory and instead took pre-orders for styles, which meant customers would sometimes wait three-plus weeks for a product to be delivered. Each was made in China, shipped to Choosy’s Missouri-based warehouse and then delivered to customers’ doorsteps. The company quickly realized that, in the age of Amazon and next-day shipping, it needed to move faster.
When Choosy relaunched in November of last year, building up enough inventory ahead of launching a new product was a big focus, to avoid selling out and keeping customers waiting for deliveries for weeks. Now, for each of its monthly drops, it makes 80 units of each item upfront.
“Then, as it starts to sell out, we will then restock thousands, and then sometimes tens of thousands of units, depending on [demand],” said Jessie Zeng, co-founder and CEO at Choosy.
“Delivery promise is extremely important for businesses of any size, because this is the first opportunity an online brand has to build trust and loyalty with their customers. For young and new brands, transparency in the delivery process can truly make or break that brand-customer relationship,” said Arpan Podduturi, director of product for Shopify.
Choosy products will often start to sell out on the day they launch. The company promotes new products on Instagram to its 121,000 followers and through unpaid partnerships with influencers. In those instances, Choosy immediately puts in an order to its manufacturer for more units. The company’s restock timeline is currently about two weeks, followed by another week to clear customs and reach customers. The company is both China- and New York-based, but manufacturing takes place overseas as Zeng’s family business is in textile manufacturing in China.
Artificial intelligence is a big part of how the company figures out what styles are trending, and therefore, what dresses or tops to design and sell. It built a proprietary AI tool to scan Instagram to find what outfits customers like and would want to buy, but for a fraction of the cost.
But there is a human component to it. Customers are encouraged to comment with the hashtag #GetChoosy under Instagram posts featuring pictures of outfits that they love worn by celebrities or influencers. The technology then learns what styles are most popular across Instagram and among Choosy customers, based on that user-generated data. Using that data, along with help from some “style scouts,” or merchandising and design experts on the team, Choosy is designs its next drop.
Now, as the company has grown and learned more about what customers like and what styles are trending, Choosy is working to improve its technology to make it more efficient.
“We’ve built out this computer vision technology to start grouping these outfits together,” said Zeng. Computer vision technology, which works with AI, is the science behind getting computers to see, identify and process images.
“AI technology on computer vision currently is only able to tell things like whether an item is a dress or a shirt. We’ve trained the data to be able to say that it’s a shirt, it’s leopard print and it has a deep v-neck, and we can share that and all the other trending things picked up by our algorithm that are in the same vein,” she said.
This method gives the company’s merchandising and design teams a lot of data to work with. It scans some 95 million Instagram posts a day to pinpoint the most popular styles. The company can, for example, use its technology to analyze all the leopard-print, V-neck shirts being posted to Instagram and figure out what the optimal design will be for the Choosy customer.