Wayfair: Gemini enriches a 30M+ product catalog 67% faster and lifts filter conversion by ~2%

• Wayfair: Helping customers create spaces they love with AI-powered discovery — Google Cloud (customer story), 2025
• Wayfair and Google Cloud Announce Expanded Partnership to Transform Online Retail with Gemini — Google Cloud Press Corner, 2025-01-12
• Wayfair Deploys Google AI to Scale and Enrich Product Catalogs — PYMNTS, 2025-01-12

Background

For more than two decades, Wayfair has been building "the destination for all things home": the company's online catalog counts more than 30 million products from over 20,000 suppliers worldwide — for every budget, style, and space. The company's strength — its assortment — is also its biggest operational pain. Every product listing needs accurate attributes: color, style, material, dimensions. Without them, shoppers can't find items through search and filters, and the huge selection turns from an advantage into a source of frustration.

On January 12, 2025, timed to the NRF Big Show in New York, Wayfair and Google Cloud announced an expanded partnership around Gemini, Vertex AI, and Google Workspace. The press release fixed two work fronts: enriching the product catalog with Gemini models, and boosting employee productivity via Gemini in Google Workspace — from email drafts in Gmail to document summarization and presentation templates, including custom "Gems" for specialized tasks. Google Workspace is being deployed to thousands of employees worldwide.

It is important to see the case's dynamics: it did not freeze in January 2025. The current version of the Google Cloud case study shows the project migrating from back-office tagging automation to customer-facing scenarios: a pilot of the AI design tool Muse, a visual Discover tab in the app, and co-development of the Universal Commerce Protocol (UCP) — Google's open standard for agentic shopping, where Wayfair is a foundational partner. Wayfair CTO Fiona Tan frames the ambition: "We want to take away the friction from our customers, but also look at ways that we can disrupt and supercharge what they used to do before. Generative AI and agentic capabilities allow us to take a fresh new look at what the process is and how we can optimize."

Google, for its part, positions the partnership as a showcase: "Wayfair is redefining what's possible in retail with Google Cloud's generative AI," said Matt Renner, President of Global Revenue at Google Cloud, at the announcement. "By harnessing the power of Gemini and Google Workspace, Wayfair is not only automating complex tasks and boosting employee collaboration, but also creating more personalized and engaging experiences for every shopper."

For the market, this is one of the most instructive retail deployments of generative AI: it starts not with a flashy chatbot but with the boring, foundational task of catalog data quality.

Problem

Manually tagging attributes for a 30M+ item catalog is slow and expensive, and cannot keep pace with the constant inflow of new products from thousands of suppliers. Attribute gaps and errors hit revenue several times over: shoppers fail to find items via filters, wrong dimensions in listings drive returns, and under-described products simply drop out of recommendations.

The other half of the problem sits on the shopper's side. Buying furniture is a high-consideration decision: it requires trust and minimal friction. Yet customers often can't phrase their query in professional terms. "Part of the problem is our customers can't actually say that they want a 'mid-century sofa,'" says Fiona Tan. "They might know that they want it to be green, but that's the end of it." Traditional online shopping with manual page scrolling quickly becomes overwhelming: a shopper might be shown hundreds of different green sofas yet find it hard to narrow the options, because they can't name the style they have in mind.

These two problems are interlocked: as long as catalog attributes are incomplete and inaccurate, no "smart" search or AI recommendations can work — there is nothing for them to stand on. Shopping for the home, as the company itself stresses, relies on building trust and reducing friction — and trust collapses at exactly the moment a delivered sofa doesn't match its listed dimensions. Finally, scale makes the task fundamentally unsolvable by people: with 30 million products, even a small share of monthly new arrivals is a tagging volume no content team could be hired for. Wayfair needed a way to enrich the catalog automatically, fast, and at a quality sufficient to build the entire customer experience on this data — from filters to design tools.

Solution

Wayfair's first step was moving catalog enrichment to Gemini models on Google Cloud (on the Gemini Enterprise Agent Platform in the current case study version; on Vertex AI in the January 2025 press release). Instead of manual tagging, Gemini automatically categorizes products and fills in attributes like color, style, and subject. The effect is twofold: Wayfair can onboard items quickly and make them available to customers with speed — while providing listings with as much rich, accurate detail as possible. "At the end of the day, we're trying to matchmake the best products for our customers," Tan explains. "Making sure the product information is as complete as possible, we are enriching it, is super important. With the capabilities that we now have, we can do that at scale."

Gemini's multimodality opened adjacent scenarios that were not in the original plan: automatically catching errors in listed product dimensions from photos, and flagging inappropriate materials in listing content. This is a typical pattern of mature deployments: the main use case pulls in "side" ones, each of which pays for itself — a dimension error on a sofa almost guarantees a return.

The Google Cloud case study states the sequence explicitly: Wayfair's "first step" was to significantly enhance its product catalog by automating data enrichment, and that move "provided a strong foundation for advancing AI-powered discovery" that is more intuitive and personalized. In other words, the order of work here is not an accident but the program's deliberate architecture: data first, then customer products built on it.

On top of the enriched catalog, the company is building a new product discovery layer. Wayfair is piloting Muse — an AI design tool where a shopper types a query of any specificity, from "moody 1920s style living room" to just "dining room," and gets photo-realistic room imagery along with shoppable suggestions from the real catalog. The app gained a Discover tab with visual search. The company is exploring how Gemini-family models, including Nano Banana, help customers visualize products and home in on their style. "Now, with generative AI and the multimodal capabilities that come with Gemini, I can show you all the different styles and as a customer interacts, we can then arrive on a style that you want," Tan says. "You can find that very visual, inspirational journey that honestly just did not exist before." Her formula for the end goal: "Everybody should feel like they have an interior design co-pilot with them the entire time that they're shopping on Wayfair."

The fourth element is agentic commerce. Wayfair is a foundational partner co-developing the Universal Commerce Protocol (UCP) — Google's open standard establishing a common language between consumer surfaces, businesses, and payment providers. UCP is set to let customers discover Wayfair products, research them, and check out directly from AI Mode in Google Search and the Gemini app — without leaving Google, while the retailer remains the merchant of record. "The reality is our customers are shopping on our platforms, but they're also now using AI platforms," Tan explains. "You want to be able to provide the capabilities and consistent experience that allows them to discover our products on the AI platform." And she stresses: "The fact that the merchant of record is still the retailer is extremely key because we've all built on our brand promise."

The fifth front is internal: Gemini in Google Workspace for thousands of employees (drafting and responding to emails in Gmail, summarizing and proofreading documents, presentation templates) plus custom Gems for specialized tasks.

Result

The published numbers focus on the catalog front. Per the joint Wayfair–Google Cloud press release: the time needed to curate new and update existing product listings dropped by 67% across a catalog of more than 30 million products. Improving the accuracy of product attributes (color, subject) and the coverage of these tags in the catalog improved conversion rates when customers use filters by ~2%. The company estimates savings from dropping manual tagging at hundreds of thousands of dollars per year.

Framing that matters. All figures are Wayfair's estimates published by Google Cloud — the vendor partner; there is no independent audit. The ~2% metric applies to conversion in filter-driven journeys, not sitewide revenue. The dollar savings are given as a "hundreds of thousands" range without an exact amount. The difference in wording between documents is also telling: the press release says curation time was reduced "by 67%," while the later case study version is more cautious — "up to 67% faster." Finally, there are no public metrics at all yet for Muse, the Discover tab, or UCP — those are pilots and announcements, not measured results.

Editorial analysis. First: this case is a rare, clean example of a "data before storefront" strategy. Wayfair first fixed the foundation (catalog attributes) and only then started building customer products on top — Muse, visual search, agentic commerce. The reverse order — a flashy AI facade first, data later — is what usually produces disappointing pilots. Second: a ~2% conversion lift on filters across a 30M-product catalog is the classic economics of large funnels, where small percentages mean more than loud demos; that said, the absolute dollar impact cannot be assessed because Wayfair does not disclose the share of filter-driven sessions in sales.

The third observation concerns UCP: Wayfair's bet on a protocol where the merchant of record stays with the retailer is an attempt to enter the agentic shopping era without ceding margin and customer relationships to the platforms. If agentic commerce becomes a meaningful channel, catalog attribute quality — where this case began — will be the entry ticket: an AI agent needs machine-readable, accurate product data even more than a human does. In that sense, the "boring" catalog tagging looks like the most far-sighted part of the whole program. Industry background confirms the direction: per PYMNTS Intelligence data cited in coverage of the case, 92% of companies use AI-driven personalization for growth and 77% of business leaders rank generative AI as the most impactful emerging technology — the question has long been not 'whether' but 'in what order.'

Lessons learned

1. Data before storefront: Wayfair started not with a shopper chatbot but with catalog attribute quality — the foundation under search, filters, recommendations, and every subsequent AI product.
2. A ~2% conversion lift on filters across a 30M-product catalog is real money: small percentages on large funnels often beat flashy pilots.
3. Multimodality pays off in side scenarios: catching dimension errors from photos and flagging inappropriate content were unplanned benefits that reduce returns and risk.
4. Measure impact as 'speed + quality' together: 67% faster only matters alongside better attribute accuracy — otherwise you accelerate garbage production.
5. A quality catalog is the entry ticket to agentic commerce: AI agents (UCP, AI Mode) need machine-readable, accurate attributes even more than humans do.
6. In agentic channels, defend the merchant-of-record role: Wayfair co-develops the standard precisely so that checkout on AI surfaces doesn't strip the retailer of margin and customer relationships.
7. Vendor-published figures without independent audits should enter your own business case with a discount and a mandatory in-house A/B test — especially since even the vendor's wording softens over time ('by 67%' → 'up to 67%').

Frequently asked questions