https://www.gstatic.com/bricks/image/b43d15a9-d1fa-4c9f-915e-1bc6f8113290.png

Etsy: Connecting 90 million buyers with special items using gen AI and "algotorial curation"

Google Cloud Results

~80x increase in listings per theme with "algotorial curation"
90 million personalized shopping experiences for 90 million shoppers
5% increase in visits and 3% conversion lift with SEO optimizations

Etsy creates personalized experiences for its 90 million shoppers by using Vertex AI, BigQuery, Dataflow, and Gemini models to better understand inventory, customer intent, and individual buyers.

Leveraging AI to amplify human connections

There's only one marketplace where you know you're buying from another real person on the other side, and that's Etsy. We're not trying to replace that human connection. We want to amplify and scale that connection with AI.

Mahir Yavuz

Senior Director of Engineering for Search, Recommendations, and Ads at Etsy

"Keep commerce human." That's the mission of Etsy , the online global marketplace that's home to a universe of special, extraordinary items, from unique handcrafted pieces to vintage treasures. Items on Etsy are made, handpicked, or designed by a real person, and every transaction is a true human-to-human connection.

Etsy is working to scale that human connection so that each one of the more than 90 million individuals who shop on the marketplace can easily find the items they consider special. With an inventory of more than 130 million items offered by more than 5 million sellers — and changing daily — Etsy turned to Google Cloud for data and AI solutions to help.

"There's only one marketplace where you know you're buying from another real person on the other side, and that's Etsy. We're not trying to replace that human connection," says Mahir Yavuz, Senior Director of Engineering at Etsy. "We want to amplify and scale that connection with AI. AI offers many opportunities to scale fast, which is our main goal, because we want to connect buyers with the right sellers as quickly and easily as possible."

The first step in this journey was to create a foundational dataset using Gemini models. The team then used Google Cloud AI technologies, with Vertex AI and BigQuery at the center, to connect, verify, and validate this data. With a core dataset in place, Yavuz and his teams turned their focus to better understanding Etsy inventory, visitor intent, and each of the marketplace's 90 million buyers.

How Etsy crafts human connections to unique goods using AI

Transforming inventory data for more relevant and delightful shopping experiences

With more than 130 million items for every taste and occasion, it's no small feat to match buyers with exactly what they're looking for. Etsy supports that search and discovery by working to deeply understand the inventory available on its marketplace.

"I like hats," says Yavuz. "I wear hats all the time. On Etsy, we have a lot of hat shops. You can find handmade hats. Vintage hats. Custom hats." No matter the item, Etsy tries to glean as much detail about it as possible, determine the best categorization and taxonomy for the item, and assess its popularity. In the process, Etsy's systems generate, evaluate, enhance, and regenerate lots of data, seeking to improve accuracy and precision to quickly surface the ideal item for a shopper.

Our inventory comprises more than 130 million items from more than 5 million sellers. Google gen AI and LLM technologies help us enrich our data to better understand our inventory's distinctive characteristics.

Mahir Yavuz

Senior Director of Engineering for Search, Recommendations, and Ads at Etsy

AI enables this deep understanding of Etsy inventory at immense scale. "Some of these things were not possible before the invention of LLMs," says Yavuz, "because it would cost so much money and time. Gen AI is fast, it's scalable, and it's also affordable now."

AI also helps Etsy identify items creating cultural buzz. Trends often emerge on social media, and Etsy sellers quickly spin up and list items referencing these trends. Even as the trend is still taking shape, models like Gemini help Etsy merchandisers quickly identify an emerging trend and classify a related new item.

Most of Etsy's inventory management products are built on Google Cloud and use BigQuery, Dataflow , the Vertex AI Platform, and Gemini models to understand inventory and connect every buyer with the right item. "Our inventory comprises more than 130 million items from more than 5 million sellers," says Yavuz. "Our approach to inventory search and discovery requires dynamic understanding or insights. That's only possible with today's generative AI technologies. Google gen AI and LLM technologies help us enrich our data to better understand our inventory's distinctive characteristics."

Understanding intent to quickly show shoppers items that "click"

With the use of Gemini, Etsy improved alt text generation for listings on Etsy, increasing SEO-driven visits by 5% and boosting conversions by 3% for our sellers.

Mahir Yavuz

Senior Director of Engineering for Search, Recommendations, and Ads at Etsy

When a shopper drops by, the first thing Etsy needs to figure out is why they've come. While some intent insights may be available — such as past purchases, items liked, and items favorited — determining what they're looking for can be much more difficult — after all, what if the buyer themselves isn't sure what they want?

"In the case a buyer is unclear what they want," says Yavuz, "we pick the best items for you using AI." Etsy does this in two ways: First, a team of experts creates collections of listings designed to exemplify a certain trend or style. Then Yavuz's team amplifies that work. "Our experts cannot create millions of different styles or listings, but they can create the seed dataset, which we can scale with AI," explains Yavuz.

Second, Etsy suggests "pathways" buyers can use to explore marketplace inventory. These pathways are built with the help of Vertex AI using advanced machine learning techniques to find semantically relevant objects. This approach surfaces the best of Etsy as quickly as possible, in response to each visitor's choices.

Intent becomes clear as buyers search. For this reason, the Etsy team works to give buyers an excellent search experience. "It doesn't matter whether the buyer is looking to find something only on Etsy," says Yavuz. "It's also important to give them the best of the information anywhere on the web." Google, for example, shows lots of alt text information in its search results. "With the use of Gemini, Etsy improved alt text generation for listings on Etsy, increasing SEO-driven visits by 5% and boosting conversions by 3% for our sellers," says Yavuz.

Enriching Etsy buying and selling experiences with multimodal AI

When Yavuz and his team first started testing AI in inventory management, they were excited by the high quality of data presented in specific formats. It got them thinking about the many other things they might do with that same data quality at scale, which led to additional inventory management innovations for Etsy.

More recently, multimodal data and AI has gained importance. "Our sellers provide us images, videos, and textual information, and we want to use everything possible," says Yavuz. "The ability to go multimodal has become critical to providing great shopper experiences."

The ability to go multimodal has become critical to providing great shopper experiences.

Mahir Yavuz

Senior Director of Engineering for Search, Recommendations, and Ads at Etsy

Making "discovering special" easier to benefit both shoppers and sellers

With Google Cloud, Etsy is crafting a new way to discover special.

Mahir Yavuz

Senior Director of Engineering for Search, Recommendations, and Ads at Etsy

Presenting each buyer on Etsy with items meaningful to them means discovering their taste and style. This involves a search component and a discovery component.

With search, Yavuz says results should reflect the individual buyer's price point and tastes. Constructing this type of search experience requires understanding shoppers as much as possible — and thus necessitates the creation of ultra-enriched metadata.

Similarly, Etsy uses data to shape buyer discovery. "When you get to your Home feed — one of our new experiences — you may see content grouped into collections like trends," says Yavuz, "Those trends are personalized for you."

Let's say, for example, a shopper sees a collection titled "Island Luxe," featuring tropical-inspired listings across categories. Which trends appear first, and the order of the items within them, are unique to that shopper. "We use AI to amplify that collection, which is based on our merchandising experts' seed collection. We find similar listings, show them to you, and then use Gemini models to understand your preferences and continue to customize your feeds and personalize your experience," says Yavuz. This curation at scale boosts listings per theme by 80x to show buyers a world of possibilities.

Various content surfaces across the Home experience are also specific to each buyer. "Again, we initially curate these, but then we use gen AI to understand your profile and present tailored merchandise selections that we think you'll love," says Yavuz.

Whether considering inventory, intent, or individual buyers, Yavuz finds Google Cloud data and AI technologies transformational. "With Google Cloud, Etsy is crafting a new way to discover special."

Etsy is the global marketplace for unique and creative goods that connects makers, artisans, and creative entrepreneurs to millions of buyers looking for something special. Home to a universe of special, extraordinary items, from unique handcrafted pieces to vintage treasures, Etsy offers an inventory of more than 130 million items from 5 million sellers and serves more than 90 million customers worldwide.