Customer Success Story

HangTen How a lean DTC team influences 23% of sales with fashion-native AI

Young couple sitting close, woman resting her head on man’s shoulder, both in casual sweatshirts against a bright blue background.

23%

of total sales influenced by YesPlz AI search and recommendations

+49%

search-attributed sales within 3 months of the new search

+11.8%

search conversion with Learn to Rank, in an ongoing live A/B test

6

YesPlz AI Tools Leveraged

Snapshot:

Rather than simply providing a solution and leaving us to manage it, their team actively reviews operational results, suggests improvements, and works closely with us to find better ways to engage customers and increase conversions.
Ryu Hansun
eCommerce Manager, HangTen

HangTen once wondered whether a catalog its size really needed AI. But shoppers expected personalized discovery, AI assistants started recommending brands, and the lean ecommerce team couldn't keep up by hand.

So HangTen started small with YesPlz AI: tagging and recommendations first. Next came GEO for every product, then fashion-native search. Each step reused the same product data, so the team never had to start over. Today YesPlz AI influences 23% of HangTen’s sales.

Woman modeling black hoodie and wide denim jeans, with pop-up tags suggesting related items and styling options on a light blue background.Key Challenges & Results

  1. Lean team, growing catalog.

    Challenge: Tagging products by hand couldn’t keep up as the line expanded. Searches came back empty even when the right product was in stock.

    Result: Every product is tagged automatically from its image, and no-result searches dropped to under 1%.

  2. No time for personalization.

    Challenge: Basic recommendations left shoppers uninspired, and nobody had the hours to curate looks by hand.

    Result: AI recommendations now influence 13.37% of total sales, and a redesigned Complete the Look bundle grew its sales contribution 3.7x.


    3. Search that never stays done

Challenge: Shoppers search in many different ways, and trends change every week. Synonyms and popular searches needed constant manual updates.

Result: Search-attributed sales grew 49% in 3 months, and Learn to Rank lifted search conversion 11.8% in a live A/B test.


4. Invisible to AI assistants

Challenge: Product pages carried only basic data, so search engines and AI assistants couldn’t fully understand or recommend them.

Result: Valid items in Google’s Rich Results Test went from 8 or fewer to 20, and HangTen is already getting shoppers and orders from AI assistants.

Solutions, in Three Phases

Phase 1 · Discovery that runs itself

YesPlz AI tagged every HangTen product with detailed fashion attributes and keeps the tags up to date as the catalog changes. On top of that data, four recommendation widgets went live with no manual curation.

  • You Might Also Like

  • Frequently Bought Together

  • Complete the Look

  • Cart Recommendations



Result: AI recommendations now influence 13.37% of total sales.


Tuned after launch: Complete the Look, 3.7x

Complete the Look first showed several separate styling items, and it underperformed. Reviewing the data, the YesPlz team noticed HangTen’s shoppers like to get in and out quickly. So we suggested one clear bundle with a simple bundle UI instead. Its sales contribution grew 3.7x.

Phase 2 · GEO for every product

YesPlz AI turned the same product data into content that AI assistants and search engines can read: descriptions, keywords, structured attributes and internal links. It updates automatically for every product, new arrivals included.


Result: Every product is GEO-ready, and HangTen is already getting shoppers and orders from AI assistants.

Phase 3 · Fashion-native search that tunes itself

Hybrid search understands fashion language, not just keywords. Learn to Rank orders results by what shoppers actually click and buy. Popular searches and 154 synonym rules update automatically.


Result:

• Search-attributed sales: up 49% within 3 months

• Search conversion: +11.8% with Learn to Rank, in an ongoing live A/B test

• Searches leading to a product click: 39.3% → 58.6%

• No-results searches: less than 1%



See It in Action: Before and Now

AI Product Tagging
Every product tagged from its image, automatically and kept current

  • Before: tagged by hand, couldn’t keep up

  • Now: automatic, always current

Screenshot of a fashion tagging tool showing a woman in a cream fleece jacket beside auto‑generated clothing attributes and confidence scores.

Style-based Similar Recommendations

The #1 recommendation driver, no manual curation

  • Before: only “customers also viewed” picks, no style matching

  • Now: style-matched alternatives, no manual curation

Online store carousel showing four recommended men’s dark gray denim pants with prices and discounts in Korean won.
Bundle Outfit Recommendations
One clear bundle, 3.7x sales contribution

  • Before: several separate styling items

  • Now: one clear, shoppable bundle

Online shopping page showing bundle recommendations for black pants, a beige graphic hoodie, and a black sweatshirt with Korean text and prices in won.Shopping Cart Recommendations
Style-matched add-ons right before checkout

  • Before: no recommendations in the cart

  • Now: add-ons that match what’s already in the bag

Online product shopping cart recommendations showing a green argyle cardigan and three styles of loose wide denim pants worn by models.Fashion-Native Search

+49% search-attributed sales in 3 months, less than 1% no-result searches

  • Before: keyword search with empty results pages

  • Now: under 1% no-result searches, with synonyms managed automatically

Online fashion-native search showing search results for “fleece,” with four young models wearing different zip-up fleece jackets.GEO-Ready Product Pages

From 8 to 20 valid items in Google’s Rich Results Test

  • Before: 8 or fewer valid items, missing the details search engines and AI assistants need

  • Now: 20 valid items, with structured product, merchant and review data, updated automatically
    Google Rich Results Test showing a URL with 20 valid items detected and crawl completed successfully on November 17, 2026.

Lessons Learned

  • Start with the data layer. Tagging came first, and recommendations, GEO and search were all built on it. No phase had to start from scratch, because the product data was already there.

  • Small automated wins add up. No single widget changed the business on its own. Together, search and four recommendation surfaces now influence 23% of sales.

  • Launch is the start, not the finish. After reviewing Complete the Look's results together, switching to one clear bundle grew its sales contribution 3.7x.

  • Lean teams need tools that maintain themselves. Synonyms, popular searches and GEO content update automatically, so results keep growing without adding hours to the team's week.

Since implementing YesPlz AI, we have seen significant improvements in how customers discover and explore products on our D2C store. Our search sales attribution grew by 49%, and product recommendations now account for an additional 13.37% of total sales. Beyond these results, what we value most is how YesPlz AI continuously works with us to improve the overall shopping experience. From improving search accuracy and helping customers find relevant products more easily to enhancing product recommendations and styling suggestions through Complete the Look, their solutions have become an important part of our daily store operations. YesPlz AI is more than just a technology provider to us. They are a reliable partner who understands our business and grows alongside our team.
Ryu Hansun
eCommerce Manager

Behind the Curtain

Curious to see how the all-in-one discovery solution works for you?

FAQs

How is "23% of sales" measured?

With last-touch attribution: each order is credited to one source only, so search and recommendations never double-count. The figure covers all time since go-live.

Do we have to launch everything at once?

No. HangTen started with tagging and recommendations, then added GEO and, later, fashion-native search. You can start with a single feature, such as GEO or search, and add more as you grow. Our team will help you put together a package that fits your goals.

Do we need engineers to set it up?

Nope! HangTen didn’t need a single engineer. We took care of the whole setup on their Cafe24 store and styled every widget to look right at home with their design. All HangTen’s team had to do was review and approve.

How long until we go live?

Typically 2–4 weeks from kickoff. HangTen's first phase went live in 2.5 weeks.

Will we get regular performance updates?

Yes. Every month we send a report with your key performance metrics, your top search queries, and the synonyms we’ve added. We also flag what to improve next. HangTen’s Complete the Look bundle came out of exactly this kind of review.

Is there an extra charge to add new features?

Most features are included in your plan and can be turned on at no extra cost. Custom features built for your store, like a new widget or a tailored integration, are quoted by scope before any work starts. No surprises.

More Case Studies