← ALL CASE STUDIES
Case Study · Retail & Jewelry

Every search now ends somewhere.

Nobody can describe a ring. Shoppers arrive holding a screenshot — from Instagram, from Pinterest, from an ex’s wishlist — and the search box demands words they do not have. This retailer let the photo BE the search. Zero-result pages went extinct.

0%Zero-result searches
<1sPhoto to inventory matches
2.3×Search-to-cart conversion
Jewelry e-commerce retailer
IndustryRetail & Jewelry
Catalog8,000+ SKUs · fine & fashion
Built byPromata — Blueprint → AI Build
StatusLive · matching right now

Does this sound like your store?

Customers describe products with their hands — and your search box wants keywords.
“Do you have something like this?” arrives with a screenshot, daily, in your DMs.
Your zero-results page is one of your most-visited pages. It is also an exit.
Shoppers search “gold twisty ring thing” — your catalog says “Aurelia Torsade, 14k.”
The product they wanted was in stock. They just could never have found it.
IF YOU CHECKED THREE OF THESE — THIS CASE STUDY IS ABOUT YOU.
The Challenge

The inventory was rich. The search box was the doorman.

Jewelry is bought with the eyes and searched with the wrong words. Shoppers arrived inspired — holding an image of exactly what they wanted — and were asked to translate it into catalog vocabulary they had never seen. Most did not try twice.

⚠ TEXT-SEARCH STOREFRONT
“rose gold twisty ring”0 RESULTS · EXIT
Screenshot in the DMs, awaiting a humanANSWERED TOMORROW
Shopper browses 11 pages hunting one styleGIVES UP, PAGE 4
Merch team writes ever-more keyword tagsNEVER ENOUGH
The exact ring, in stock, unfoundINVISIBLE INVENTORY
✦ VISUAL-SEARCH STOREFRONT
Same shopper uploads the screenshotMATCHES IN <1S
Top matches ranked by visual similarity94% · 89% · 86%
“Similar styles” on every product pageBROWSING BECOMES FINDING
New SKUs indexed on upload, automaticallyZERO TAGGING
Every search ends on a product0% DEAD ENDS
VS

THE SAME SHOPPER, THE SAME RING — A VOCABULARY TEST VS. A MATCH IN UNDER A SECOND

  • Vocabulary was the conversion killer

    The catalog spoke jeweler; shoppers spoke feelings and screenshots. Every failed translation was a sale walking to a marketplace with better search.

  • Inspiration is visual and impatient

    Social media hands shoppers a picture, not a product name. The store that can accept the picture wins the moment; the one that cannot, funds a competitor’s ad.

  • Tagging could never keep up

    Human keyword tagging is finite; the ways people see a ring are not. The gap between them was the zero-results page.

  • Dead ends taught shoppers to leave

    A zero-result search does not just lose one sale — it teaches the shopper this store cannot find things.

The Solution

The photo is the query.

We built visual search into the storefront: shoppers upload any image — screenshot, photo, saved pin — and the engine understands what it sees, then matches it against the live catalog by visual similarity. Three panels tell the whole story:

1 · What the shopper has
A SCREENSHOTNO WORDS
2 · What the engine sees
ROSE GOLDOVAL CUTPAVÉ BAND~1.2 CT4-PRONGVINTAGE PROFILE
3 · What the shopper gets
AURELIA · 94%VESPERA · 89%DUETTA · 86%IN STOCK · ADD TO CART
  • Understands jewelry, specifically

    Not generic image matching — the engine reads metal, cut, setting, stones, and silhouette the way a jeweler’s eye does, so matches feel right, not merely similar.

  • The catalog indexes itself

    Every product photo is understood on upload. No tagging project, no merch backlog — an 8,000-SKU catalog became fully visually searchable without one keyword written.

  • Similar-styles everywhere

    The same engine powers “you may also love” on every product page — real visual similarity instead of “people also bought,” which quietly lifts every session.

  • Works where shoppers already are

    Upload on the site, or send the photo by DM — the store meets the screenshot wherever it lives.

Results & Business Impact

The dead end became a front door.

Photo searchesEVERY ONE RETURNS MATCHES
Shoppers who engage a match3 IN 4
Search-to-cart conversion2.3× TEXT SEARCH
THE VISUAL-SEARCH FUNNEL — FROM SCREENSHOT TO CART, NO VOCABULARY REQUIRED
0%Zero-result searches
<1sPhoto to matches
2.3×Search-to-cart vs text
8,000+SKUs, self-indexed
  • Zero-result searches: eliminated

    Every photo search returns ranked, in-stock matches. The exit page became an entry point.

  • Visual searchers buy more than twice as often

    Shoppers who arrive with a picture arrive with intent — the engine finally lets that intent land on a product.

  • Invisible inventory became visible

    Styles that no keyword ever surfaced now sell — the long tail of the catalog joined the store.

  • The DMs got easier too

    “Do you have something like this?” now has a one-second answer the team can send with a link — or the shopper finds it herself.

Your customers have been showing you what they want all along.

The screenshots were always arriving — in DMs, in emails, described helplessly over the phone. The store just could not accept them as input. Now it can, and the gap between inspiration and checkout is one upload wide.

0%Zero-result searches
<1sPhoto to matches
2.3×Search-to-cart

What we didn't solve

The search matches what is in stock. A photograph of something the store does not carry returns the nearest real alternatives rather than nothing — but it cannot conjure inventory, and a shopper looking for a discontinued piece still leaves without it.

Can YOUR store accept a screenshot?

If your customers see products they cannot search, tell us about your catalog. We’ll show you the photo-as-query version, before you spend a dollar.

Book a discovery call

Want results like these?

Tell us where your team is losing hours. We'll show you exactly what automation can do about it — and what it's worth, before you spend a dollar.

30 minutes · No pitch deck · An honest first read

Case study: reverse image search for jewelry commerce