Shopping for clothing online has liberated us from the need to brave the endless aisles, fluorescent lights and sale-hungry crowds of the brick-and-mortar retail inferno. But anyone who has found themselves two hours deep into a fashion rabbit hole, with nothing to show for it but 15 open tabs, four full shopping carts, an earful of YouTube clothing haul reviews and the gnawing anxiety of the overwhelmed, shopping online can feel like a chore.
Enter Blend, a U.K.-based startup that is using AI to cut through the noise and help shoppers find personalized product recommendations to suit their style, budget and size.
"The vast majority of retailers do absolutely no personalization, and in the instances when they do, they only personalize according to historic purchase data," Blend co-founder Jemima Bunbury told TechCrunch. "When trends are changing relatively quickly, and people's style does change over the course of their lives, it doesn't stay relevant for a user to have such historic recommendations."
Blend participated in TechCrunch Disrupt 2023 as one of the Startup Battlefield 200 companies. At the event, the startup launched its MVP -- an app that will slowly open to the 2,000 users on Blend's waiting list. After raising angel investment in April, Blend is now on the hunt to secure investors for its seed round. The startup will use those funds to build out additional features on the app and push for a full-scale launch.
Blend has already signed on over 250 retailers, including Net-a-Porter, a luxury retailer. The startup's go-to-market strategy targets users aged 18 to 34, "very digital, native mobile-first shoppers" who are starting to define their personal style as they accrue disposable income. Blend is launching in the U.K. first and then hopes to move into the U.S. market.
"We hope that by attracting first the very fashion-forward, trendsetting crowd, we can then move more mainstream from there, but it's much more difficult to go the other way round," said Bunbury. "Ultimately, the vision is really to be the front door for every online shopping experience, and therefore, to be the largest-scale retailer because of that ability to personalize and only present people with the 1% of the internet that is most relevant to them."
Generative AI we can get behind
Blend co-founders Bella Levin (left), Jemima Bunbury (middle) and Eva Piskova (right). Image Credits: Blend
The fashion industry has tapped the generative AI frenzy in a range of ways. Some companies are using natural language processing algorithms to improve the customer service experience. Others are using image generation to create new designs. There are also applications in production improvement, trend forecasting, inventory management and virtual try-ons.
Blend's approach centers around transformer technology and recommendation algorithms, powered in large part by user interaction data. Transformer technology, which makes up the tech stack of popular generative AI models like ChatGPT, is a model for teaching computers how to understand and generate human language. In the world of fashion, this means it can better understand user preferences and make tailored clothing recommendations.
"The main thing that's always important when it comes to AI is what data you are actually putting into [the model]," said Bunbury, noting that the founding team decided on an app rather than a web page in part because it's easier to track a user's data that way.
When the user opens the app, they'll scroll through a feed that is a mix of product imagery and descriptions that have been pulled from different retail and e-commerce sites. Their feed will also feature short-form videos and product curations from influencers who can earn an affiliate commission on any sales they generate.
As the user scrolls, Blend collects data on how they interact with the app, whether they're liking products, saving them, sharing with a friend, "or simply how long you're looking at one product," according to Bunbury. Blend uses all of that data to form a picture of the user, who has already pre-set preferences to size and budget. The more a user interacts with the app, the more personalized their recommendations will become.
On the back end, Blend is comparing products and users to get a statistical picture of which products will be right for which users. So, for example, let's say there are two users who were actively using the app three months ago. User A pauses engagement with the app, while User B continues to engage regularly, and sees her feed adjusted according to new trends. Rather than let User A's recommendations stagnate, Blend will use User B's data to inform recommendations to User A.
"By tracking those cultural trends and how different people's styles are similar or different, we can use that data to inform other people's recommendations," said Bunbury. "So the personalization gets more powerful the more users we have on the platform to base those off of and create cohorts."
The AI model behind the app is impressive not only because it can recommend you the right outfit today, but also tomorrow, next week, next year. It's dynamic, and it tracks how a user's style changes over time.
Blend also helps users find the right fit for their body type, something retailers who have to go through expensive returns cycles appreciate, as well. Part of getting this right is allowing the user to set their preferences for what their size is for different body parts and determine what their body type is. But that information isn't always reliable -- brand sizing charts can differ wildly and most of us aren't good at classifying our own body shapes.
That's where the user-generated content enabled by the app kicks in again. The hope is that users will take photos of themselves in their new clothes and post them on the app, giving Blend's AI engine and other users a diverse representation of what specific products look like on different frames. Down the line, Blend hopes to incorporate reviews and a voting system to help users better determine the right size for them.
The more a user engages with their Blend feed, the better the personalized recommendations become. Image Credits: Blend
The three moving parts in Blend's business model are: 1) Shoppers; 2) Influencers; and 3) Brands.
Blend is predominantly trying to solve a user problem, but to do that, it needs to partner with influencers and brands, both of which stand to gain, as well. By partnering with Blend, both influencers and brands can diversify revenue streams and appear across multiple different channels in a very light-touch way.
For brands specifically, Blend could present as a powerful market marketing platform.
"For most brands, the key difficulty is getting your products in front of the right audience and having a risk-free way of advertising," said Bunbury. "With social media advertising, yes, you can target fairly well according to demographics and user group, but even then it isn't necessarily based on what their style is. Whereas we should have this incredibly granular style-specific dataset that will allow us to put the right brands in front of the right users when they are actively looking to be buying."
Blend wins by taking a commission on sales from partner brands and retailers, which can vary depending on retailer, according to Bunbury.
The first version of the app will link out to a brand's website to complete the transaction there. Future versions will allow users to get to the point of sale within the app for a more seamless user experience.
"There's huge growth potential just in that, but we're also aware that with the dataset we have and with our ability to put brands in front of users, there are also lots of B2B revenue lines in the future," said Bunbury. "Things like advertising, data and analytics on trends, being able to forecast what sorts of products will be selling and at what quantities."
On the consumer side, Blend says it might launch a subscription service in the future for additional premium features, like end-of-stock alerts, discount alerts or early access to brand products.