AI Visual Search for a 500,000-SKU Furniture & Interior Platform
AI Visual Search for a 500,000-SKU Furniture & Interior Platform
How a leading Japanese furniture platform operator deployed AI-powered visual search across both B2B and B2C channels — enabling interior professionals and consumers to find products by uploading a photo, instead of describing them in words.
About Client
The client is a leading Japanese furniture and interior platform operator running two complementary e-commerce channels: a B2B wholesale platform serving interior professionals (designers, architects, corporate buyers) and a B2C subscription-based platform offering premium branded furniture and appliances to consumers. Together, the platforms carry over 1,000 brands — from iconic manufacturers like Herman Miller, Vitra, and Fritz Hansen to specialized Japanese furniture makers — across a catalog of 500,000 SKUs.
Challenges
Furniture and interior products present a unique search problem. Unlike electronics or office supplies — where buyers search by model number or specification — furniture buyers search by look, feel, and aesthetic. A designer who sees a chair in a showroom photo, a magazine spread, or a client’s Pinterest board needs to find that exact product (or something visually similar) in a catalog of half a million items.
Keyword search fails for visual products
How do you describe “mid-century walnut chair with woven cord seat and tapered legs” in a search bar? Most users type “chair” and get thousands of irrelevant results. The gap between what the shopper sees in their mind and what keyword search can retrieve is enormous.
500,000 SKUs make browsing impossible
Category browsing works for small catalogs. At 500,000 SKUs across 1,000+ brands, it’s impractical. Users need a way to jump directly to visually relevant products without clicking through dozens of category pages.
B2B professionals need speed
Interior designers and architects work under project deadlines. They need to identify, source, and quote furniture quickly — often from a reference photo provided by a client. Manual catalog browsing is too slow for professional workflows.
Dual platform, one search problem
The same search limitation affected both the B2B wholesale platform and the B2C consumer platform. Any solution needed to work across both channels, with different user behaviors and expectations.
Our Approach
TPS Software deployed AI-Powered Visual Search across both the B2B and B2C platforms — enabling users to find products by uploading a photo instead of typing keywords. The system indexes the full 500,000-SKU catalog and returns visually similar results in 2.5 seconds.
Upload
User uploads a photo — from their phone, a screenshot, a reference image, or a project render.
Detect
AI detects objects within the image. Multiple items in one photo can be identified separately.
Select
User selects the target object they want to find. System isolates it from the rest of the image.
Match
Vector similarity search across 500K indexed images retrieves visually similar products.
Refine
Results filtered by color, category, brand, price, and product attributes.
AI Capabilities
Dual Platform Deployment
A key design requirement was supporting two fundamentally different user contexts — professional sourcing (B2B) and consumer discovery (B2C) — with the same underlying AI engine but adapted to each platform’s workflow.
Professional Sourcing
Interior designers and architects use visual search to match reference photos from client briefs, showroom visits, or project renders to available products across 1,000+ wholesale brands.
- Search from project renders and floor plans
- Match showroom photos to catalog products
- Cross-brand discovery for specification alternatives
- Integrated with floor plan coordination tools
Inspiration-to-Purchase
Consumers use visual search to find furniture they’ve seen in magazines, social media, hotel lobbies, or friends’ homes — bridging the gap between inspiration and purchase.
- Upload from phone camera or screenshot
- Social media inspiration → product match
- Find similar alternatives at different price points
- Subscription pricing shown alongside results
Technical Results
The deployment successfully indexed the client’s full catalog and is now live across both platforms, serving visual search queries from interior professionals and consumers across Japan.
The 2.5-second response time is measured end-to-end — from the moment the user uploads an image to the moment the server returns matched results. This includes image upload, object detection, vector similarity search across 500,000 indexed images, and result ranking. For interior professionals working under project deadlines, this means a reference photo from a client meeting can be matched to available catalog products in the time it takes to switch tabs.
Business Impact
New discovery pathway
Visual search opens a product discovery channel that didn’t exist before — users who previously couldn’t find what they wanted via keyword search now have a direct path from image to product.
Professional workflow acceleration
Interior designers can match a client’s reference photo to available wholesale products in seconds — eliminating hours of manual catalog browsing during the specification phase.
Cross-brand discovery
Visual search surfaces products across 1,000+ brands simultaneously — helping users find alternatives they wouldn’t have discovered through keyword or category browsing alone.
Scalable architecture
The system indexes 500,000 SKUs with consistent 2.5s response time. As the catalog grows, performance scales linearly — no degradation as brands and products are added.
Why This Deployment Matters
This project demonstrates three things about AI Visual Search at production scale:
Scale is not a barrier. 500,000 SKUs indexed with 2.5-second response time proves that visual search works at large-catalog scale — not just for curated collections of a few hundred products. The same architecture handles catalogs of any size.
Dual-context deployment works. The same AI engine powers both B2B professional sourcing and B2C consumer discovery — with different UI integrations adapted to each platform’s workflow. One model, two channels, one consistent experience.
Furniture is the ideal category for visual search. In categories where products are chosen by aesthetics rather than specifications — furniture, interior goods, home décor, fashion — visual search isn’t a feature. It’s the search modality that matches how buyers actually think. “I want something that looks like this” is a more natural query than any keyword string.
See Visual Search in action with your catalog
We demo with your product images — not generic samples. Upload a photo, see matched results from your catalog in real time.
CONTACT OUR EXPERT TEAMNo commitment required · Works with any catalog size
