How 9AI Cut Product-Search Time 96% for a Luxury Furniture Brand
96% Faster Product Discovery
The House of Things curates 30,000+ designer SKUs but clients arrive with mood-board images and vague briefs. Designers spent 18 minutes per search trawling Magento. 9AI deployed a CLIP-powered visual search platform that returned a ranked top-10 in under 60 seconds, recovered ₹7.5L per month in designer labour, and lifted cross-sell conversions from 8% to 26%.
96%
Search Speed
faster
₹7.5L
Monthly Labour Saved
26%
Cross-Sell Rate
from 8%
180h/wk
Designer Hours Freed
The new visual search lets us wow clients in minutes instead of spending half a day hunting for the right pieces.
— Head of Design, The House of Things
The House of Things is a Udaipur-based luxury marketplace for designer furniture, art, and lighting with 30,000+ SKUs from over 200 studios. Their sales model is consultative: clients share mood boards, architects send reference images, and designers are expected to produce curated shortlists instantly. The catalog is only searchable by keyword in Magento — useless when a client says "I want something like this" and shows you a photo.
The average shortlist took 18 minutes to assemble manually. At peak inquiry volumes, the design team burned 200 hours per week on search alone — ₹10L in monthly labour before a single recommendation was made. Slow responses on WhatsApp cost deals. Incomplete shortlists undersold the catalog. Cross-sell was almost never happening because designers had no bandwidth to surface complementary pieces.
#Before vs After: The Visual Search Impact
| Metric | Before AI | After AI | Change |
|---|---|---|---|
| Search time per query | 18 minutes | < 60 seconds | 96% faster |
| Designer search hours / week | 200 hours | 20 hours | 90% reduction |
| Monthly labour on search | ₹10L | ₹2.5L | ₹7.5L recovered |
| Search match accuracy | 70% | 95%+ | +25 percentage points |
| Cross-sell conversion rate | 8% | 26% | +18 percentage points |
#What 9AI Built
#The Visual Search Platform — 5 Components
Nightly Magento Catalog Sync
A pipeline pulls all active SKUs — product photography, attribute tags, descriptions — into the platform each night. No changes to Magento required.
CLIP Multi-Modal Embedding
Each product gets a vector that encodes visual appearance, material, form factor, colour, and era — not keyword tags. Image queries and text queries are encoded the same way.
PgVector Similarity Search
Top-10 nearest SKUs retrieved by cosine similarity in under one second from a 30,000+ product corpus. No scroll, no tabs.
Style-Affinity Upsell Layer
Complementary pieces ranked by visual style proximity surface automatically alongside the main results — replacing manual cross-sell guesswork.
WhatsApp-Native Delivery
Designers run searches and share shortlinks directly inside WhatsApp Business. Zero context switching; clients receive the shortlist without leaving the conversation.
#Questions About This Project
#Results at Month 4
96%
Search Speed
faster
₹7.5L
Monthly Savings
labour recovered
+18 pp
Cross-Sell Lift
8% → 26%
180h
Team Hours Freed
per week