ThredUp taps AI search tools to give its $322M secondhand apparel business a boost
Secondhand apparel marketplace ThredUp has launched a trio of AI-powered tools to help customers refine their searches and seek product recommendations for queries beyond the typical keyword search.
With more than 4 million SKUs on its sprawling web store, the tools are designed to make it easier for customers to find exactly what they’re looking for, whether that’s a dress for an upcoming job interview or a Valentine’s Day date. The new features are rolling out now, and will be widely available to all customers starting Aug. 8, the company announced Monday. Previously, the tools were only available to a random segment of consumers as ThredUp tested and refined their capabilities based on user feedback. Unlike peer-to-peer resale platforms like Poshmark, ThredUp gets its inventory from consumers who mail in their unwanted clothes for the e-tailer to sell on its platform. Brands can also set up an online resale shop through ThredUp.
“What has changed is that now consumers are getting much fewer dead ends in the search experience,” said ThredUp CEO James Reinhart in an interview. “There was one point where you would search for, let’s say, Fourth of July outfits, and you’d get zero results.”
The first change from ThredUp is a search bar doesn’t rely on Boolean searches, keywords or product names to find a specific item. This functionality seemed outdated ever since OpenAI’s ChatGPT burst onto the scene in late 2022, according to Reinhart, who said that customers would prefer to write in a more conversational style. Now, ThredUp’s semantic search is capable of curating product recommendations based on descriptive text. As a result, the feature is able to understand queries for specific pop culture trends or references, such as “Brat summer.”
Another AI-focused tool lets users upload a photo or share an image URL into ThredUp’s search bar, and image recognition technology will scour its online marketplace to find merchandise that fits the requested visual.
Finally, a service dubbed Style Chat functions like a personal stylist, helping customers generate head-to-toe outfits from scratch. Like the search tool, users can request an outfit idea in a conversational style, and the bot will spit out a complete ensemble in a few seconds. If the bot’s suggestions aren’t quite what a shopper is looking for, they can manually change the parameters, including color, style and price.
A Modern Retail test of Style Chat’s capabilities demonstrated that it’s able to build outfits that fit most general requests. But it struggles with some higher level or more nuanced queries — a fairly common shortcoming of many AI-powered chatbots. For example, when asked to generate an “eco-friendly” outfit, the bot’s suggestions included products from fast-fashion retailers Shein and Zara, brands that have developed well-known reputations for textile waste. The same effect happened when Modern Retail requested “sustainable” styles.
Sustainability is a major touchstone of ThredUp’s business ethos. On Aug. 1, ThredUp released its third annual Impact Report, an overview of its ESG initiatives in 2023.
Style Chat is intended for practical applications, but at its core, it’s a chatbot. This means users can ask it more colorful questions. For example, when asked to write a haiku about wedding dresses, it wrote back, “Wedding bells chime bright, Layers of lace softly flow, Love’s tapestry glows.” This was followed by a series of emojis including a wedding ring.
Still, Style Chat is clearly programmed to keep the conversation focused on ThredUp’s marketplace. Modern Retail’s request for a haiku generated results for relevant garments, such as white dresses and tulle skirts. The same effect happens when Style Chat is faced with fact-based questions, such as, “How many planets are there?” Style Chat responded, “There are eight planets in our solar system, but let’s not orbit around that! How about an out-of-this-world outfit for your next stargazing adventure?” The bot proceeds to recommend black clothing and garments with abstract, space-y prints.
Retailers in general have embraced generative AI. Amazon’s shopping assistant Rufus aims to help customers find products on its massive web store. Ebay sellers can generate product descriptions with artificial intelligence. And Shopify merchants can use a chatbot to answer business queries.
But such tools do not always find a warm reception among customers. For example, Levi’s halted its pilot program of AI-generated models on its e-commerce site due to customer backlash in May.
A consumer survey from IBM found that just one-third of shoppers who used chatbots and virtual assistants were satisfied with the experience. And nearly 20% of shoppers said they were so disappointed they would never use them again.
“I’m not sure that the Style Chat will get as fast adoption as people using the image search, so I think that one will get more use out of the gate,” said Reinhart. But he also believes the chatbot “is more compelling in the long term” because it will help ThredUp tap into emerging trends on the internet and social media.
Ultimately, the success of ThredUp’s AI investments will hinge on the quality of the site’s product metadata, according to Tim Glomb, vp of digital, content and AI at performance marketing company Wunderkind.
“The deeper of that metadata you have — this is pink, this is cotton, this is soft, this is light this is breathable, all those attributes — when that data is available to the AI engine, it’s going to give a far, far better result for the consumer,” said Glomb.
To Juan Pellerano-Rendón, chief marketing officer at e-commerce logistics startup Swap, AI is more likely to help retailers if such tools are used to streamline operations, like trim costs, reduce headcount, and so on. “From a consumer standpoint, I’m not sure that it’s as sticky because the habits that we have in terms of how we buy are not determined by the tools that we have available at our disposal.”
ThredUp’s AI-focused investments come at a time when ThredUp and other secondhand apparel retailers have struggled with profitability. Although ThredUp’s revenue rose to a record $322 million last year, the company still recorded a net loss of $71 million. ThredUp’s stock has lost about 90% of its value since the retailer went public in 2021.
In a bright spot, the resale market is booming by some accounts. Last year, the total value of merchandise sold on resale platforms in the U.S. reached $20 billion, up 11% compared to 2022, according to a report published by ThredUp and analytics firm GlobalData. Gross merchandise value is expected to more than double to $44 billion by 2028, outpacing the wider apparel sector.
While ThredUp’s Reinhart declined to provide a specific numerical outlook or forecast regarding how the AI tools will drive e-commerce sales, he said he believes they’ll be a “massive, massive advantage” due to the site’s “wide breadth of SKUs.” As Reinhart put it, “When I go to shop at a multi-branded retailer, and they’ve got 70,000 things and I’m not sure where I should look, then I think AI can really help you.”