How to Localize a Product Title for Coupang Search (Beyond Translation)

Why translated titles underperform on Coupang

A Coupang title is not a description. It is a keyword container that has to match how Korean shoppers type queries into the search bar — and Korean shoppers type differently than English ones. A literal translation of "Stainless Steel Insulated Water Bottle 500ml — Leak Proof, BPA Free" produces a grammatically fine Korean sentence that almost nobody searches for.

The failure is usually not vocabulary. It's structure. Korean search queries tend to be short noun stacks with no particles or connectors: 스테인리스 텀블러 500ml (stainless tumbler 500ml). Marketing adjectives like "premium," "ultimate," or "leak proof" rarely appear in real queries, so translating them just dilutes your title with tokens that will never be matched.

Coupang's search also rewards relevance signals over time — clicks and conversions on a query reinforce your ranking for that query. So a title stuffed with words nobody searches gives the algorithm nothing to learn from.

• Translation gives you *meaning*. Localization gives you *demand* — the exact strings people type.
• Korean queries drop particles (은/는/이/가) almost entirely. Your title should too.
• Adjective-heavy English marketing copy is dead weight in Korean search.

Start from queries, not from your English title

Reverse the workflow. Instead of translating your title and hoping it matches, collect the Korean queries that already exist for your product category, then build a title out of the highest-value ones.

Three sources are enough to get a solid list. First, Coupang's own autocomplete — start typing your category noun and record every suggestion. Second, the related-search chips that appear under the search bar on results pages. Third, the top 10 organic listings for your main keyword: copy their titles into a document and look for tokens that repeat across five or more sellers. Repetition across competitors is a strong signal that the token is doing search work, not branding work.

If you can't read Hangul, this is exactly the manual step that stalls most cross-border sellers. Tools like SeoulRank surface Korean keyword volume and competitor title tokens with English glosses, so you can judge a keyword's value without guessing at the characters.

• Record autocomplete suggestions for your head noun — these are real, weighted queries.
• Diff the top 10 competitor titles and mark tokens appearing 5+ times.
• Note which tokens appear *first* in competitor titles — position signals importance.
• Watch for spec tokens (size, capacity, count, material) that shoppers use as filters.

The token order that actually works

Korean marketplace titles follow a fairly consistent, learnable pattern. Front-load the tokens with the most search demand, because both the algorithm and the shopper's eye weight the beginning of the string — and mobile listings truncate the tail.

A reliable skeleton: 브랜드 (brand) + 핵심 키워드 (head keyword / product noun) + 주요 속성 (key attribute: material, flavor, target user) + 규격 (spec: size, capacity, count) + 세트 구성 (pack/set info).

So a US brand selling a 500ml insulated tumbler ends up with something like: 브랜드명 스테인리스 보온 텀블러 500ml 1개 — brand, stainless, insulated, tumbler, 500ml, 1 unit. Short, scannable, and every token is a plausible query fragment. If your brand has zero recognition in Korea, consider moving it after the head keyword so the searchable noun sits first.

• Head keyword within the first ~20 characters — mobile truncation is brutal.
• Put spec numbers in the format shoppers type: 500ml, not 500 밀리리터.
• One head keyword per title. Two competing head nouns split your relevance.
• Pack/count tokens (2개입, 3세트) matter for bulk-intent shoppers.
• Avoid special characters and decorative separators; use plain spaces.

Konglish, Hangul, and Latin script: which to use

Many Korean product nouns are loanwords written in Hangul — 텀블러 (tumbler), 케이스 (case), 매트 (mat), 크림 (cream). Shoppers search these in Hangul, not in English letters. Writing "tumbler" in Latin script will not match 텀블러 in most cases, so the loanword must be transliterated correctly, not left in English.

The exception is brand names and certain technical strings. Global brands are often searched in Latin script, sometimes in both. When both forms have real demand, including both is legitimate — 나이키 Nike — rather than keyword stuffing. Model numbers, units, and standards (USB-C, IP68, SPF50+) stay in Latin/numeric form because that's how they're typed.

Be careful with transliteration variants. The same English word can have two accepted Hangul spellings, and one of them usually carries most of the search volume. Check which spelling autocomplete prefers before committing.

• Common nouns → Hangul, always. 케이스, not "case."
• Brand names → check demand for both scripts; include both only if both are searched.
• Specs and model codes → keep as typed: 15.6인치, USB-C, SPF50+.
• Test transliteration variants against autocomplete; pick the dominant spelling.

What to cut, and what Coupang penalizes

Coupang's listing rules and its ranking behavior both push toward clean, accurate titles. Repeating the same keyword to boost density tends to backfire, and unrelated high-volume keywords (borrowing a famous brand's name for traffic) risk listing suppression or removal.

Cut anything that is a claim rather than a query. Promotional language — 최고, 무료배송, 할인 (best, free shipping, discount) — belongs in the promotion module, not the title. Coupang generally restricts promotional wording in titles, and shoppers don't search it anyway.

• No keyword repetition — one instance per token is enough.
• No competitor or unrelated brand names.
• No promo words: sale, free shipping, best, #1.
• No emoji, brackets clutter, or ALL-CAPS runs.
• No unverifiable claims (medical, whitening, efficacy) — these carry regulatory risk in Korea, especially for cosmetics and food.

Test, measure, iterate

Treat your title as a hypothesis. Publish version one, then track where the listing ranks for each of the 5–10 target keywords you built it from. If a keyword you included isn't producing impressions after a couple of weeks of stable listing conditions, it's either too competitive for your current sales history or it's the wrong variant.

Change one thing at a time. Swap a single token or reorder the front of the title, then watch rank movement per keyword — not just overall sessions, which are too noisy to attribute. Daily rank tracking per keyword is what makes this loop workable; SeoulRank does that side in English if you'd rather not run manual searches every morning.

Two more practical notes. Titles interact with your category and option settings — a title token that contradicts your registered attributes can hurt filtered visibility, so keep them consistent. And revisit seasonally: Korean search demand shifts sharply around holidays, weather changes, and 명절 (traditional holidays), and a token that was dead in March can be your best keyword in September.

• Track rank per target keyword daily, not aggregate traffic.
• One variable per test; give each version 10–14 days.
• Keep title tokens consistent with your registered category attributes and options.
• Re-audit titles seasonally — demand for the same product shifts by month.