Lily AI lands new capital to help retailers match customers with products

During the pandemic, retailers were forced to embrace e-commerce. But some found that they struggled to maintain customer loyalty as consumer expectations changed and purchasing patterns shifted. As a result of formidable competition like Amazon, they discovered, customers have low patience for sites that don’t present them with what they want. According to research from the Baymard Institute, for every 100 potential customers, 70 will leave without purchasing.

That’s why Purva Gupta launched Lily AI, an AI-powered platform that connects a retailer’s or brand’s shoppers with products they might be looking to buy. Co-founded by Sowmiya Narayanan, Lily provides algorithms designed to power web store components like search engines and product discovery carousels.

Lily today announced that it raised $25 million in a Series B funding round led by Canaan, bringing its total raised to $41 million.

“Different shoppers search uniquely, making it essential for retail ecommerce brands to build the right product taxonomy to capture both common and long-tail searches,” Gupta told TechCrunch via email. “Think of your own frustrating experiences on retail ecommerce sites and receiving irrelevant results or worse, no results at all, even when the product you’re looking for is clearly carried by that retailer.”

Prior to co-launching Lily, Gupta served in various roles at Eko India and UNICEF. Narayanan brought her experience developing software at Texas Instruments, Yahoo! (full disclosure: TechCrunch’s parent company) and Box, where she was a full-stack web dev for the product Box Notes. 

Lily began life as an app for retailers to help understand women shoppers’ personal preferences around fashion. But when traction proved hard to gain, Gupta and Narayanan pivoted to build a more enterprise-focused solution packaged as a plug-in, software-as-a-service subscription product. 

Lily now retains a team of “experts” in fashion, home and beauty who help to refine product taxonomies, which are then used to train algorithms for product search and recommendations. (The group also researches and develops ways to turn product attributes like “ribbed fabric” and “minimalist dressing style” into a mathematical “language” that the algorithms can understand.) Essentially, Lily captures details on products based on traits (e.g. “style,” “fit” and “occasion”) and uses customer data from brands tied to the item attribute data to create a prediction of each customer’s affinity to attributes of products in the catalog.

Gupta acknowledges that there are other companies in the product attribution and automated product tagging spaces that rely on automation and AI. For example, Depict.ai provides a product recommendation tool that draws on data from across the internet. Black Crow AI is developing a platform to predict which products e-commerce customers will buy, while Constructor sells access to a framework that powers search and discovery for digital retail marketplaces.

Meta has also experimented with apparel attribute prediction for Facebook Marketplace, two years ago showcasing a system that could extract clothing attributes and fashion styles from photos of models on Instagram and Flickr.

But she argues that Lily is one of the more powerful options out there in terms of its configurability. Gupta also stressed that the platform is privacy-preserving to the extent it’s able to be, not using customer names, addresses or financial transaction information in favor of using anonymized user interactions on its customers’ ecommerce sites.

“The IT decision makers with whom we work are focused on the more concrete and tangible application of Lily versus being on the strategic frontlines. They are interested in the depth and accuracy of information Lily can provide; how we are training the models; and accuracy of output and confidence levels,” she said. “We win with the customization of our product to deliver on their needs and a dedicated customer success team available to take into account changes to goals or results over time.”

In any case, big-name customers have signed up for Lily’s services to date, including Macy’s, The Gap and its assorted brands, Bloomingdale’s and thredUP.

Lily is loathe to make its revenue figures public, and the 87-employee company says it doesn’t have a projection for the size of its headcount for the end of the year. Brushing aside questions about the secrecy, Gupta asserts that Lily is “well-positioned” to capitalize on new retail verticals in the coming months, even factoring in macroeconomic headwinds.

“Lily AI grew tremendously since the start of the pandemic, as the health crises rapidly intensified the retail shift to e-commerce and digital transformation,” Lily said. “We’ll use the new funding to further expand into enterprise and mid-market retail e-commerce brands across home, beauty and fashion … We also plan to extend our solution much deeper to further applications within the retail stack, as well as further a suite of rich analytics for our customers.”