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

Translation gets you a sentence; localization gets you impressions

A Coupang product title (상품명, sangpummyeong) is not marketing copy. It is a keyword container. Coupang's search engine matches buyer queries against the title, category, and attribute fields, and the title carries the heaviest weight for the exact phrases shoppers type. If a phrase never appears in your title, you are effectively invisible for it.

Machine translation optimizes for grammatical correctness. Korean shoppers do not type grammatically correct sentences — they type compressed noun stacks like "무선 청소기 핸디" (cordless vacuum, handheld). A translated title reads fine to a Korean and still ranks for nothing, because the words a translator chooses are often not the words buyers search.

So the job is not "say this in Korean." The job is: find the real search phrases, then assemble them into a title that reads naturally enough to earn the click.

Step 1: Find the demand-side word, not the dictionary word

Korean frequently has two or three valid words for the same product: a Sino-Korean or native term, an English loanword written in Hangul, and sometimes the raw English. Which one wins is purely an empirical question of search volume, and it flips category by category.

Examples of the split you will run into constantly:

• Cushion/pillow: 쿠션 (kusyeon, loanword) vs. 방석 (bangseok, floor cushion) — different products entirely, and MT will happily pick the wrong one.
• Water bottle: 물통 (multong, literal) vs. 텀블러 (teombeulleo, tumbler) vs. 보온병 (boonbyeong, thermos) — buyers search by use case, not by literal object.
• Skincare: 수분크림 (subun keurim, moisture cream) massively outranks 보습제 (boseupje, moisturizer) as a shopping query even though both are correct.
• Pet products: 강아지 (gangaji, puppy/doggie) usually outsearches 반려견 (banryeogyeon, companion dog) — casual beats formal in commerce queries.
• Adapters and cables: many still search the English spec token (USB C, HDMI) in Latin letters, so keep both forms in the title.

Step 2: Build the title from head term + modifiers, in buyer order

Korean product titles follow a fairly consistent structure that shoppers scan quickly. A workable template: 브랜드 + 핵심 키워드 + 속성 + 규격/수량 — brand + head keyword + attributes + spec/quantity.

Put the highest-volume head term where it cannot be truncated. Coupang shows titles clipped on mobile grids, so the first ~25–30 characters do most of the persuading. Front-load the phrase that matches the query, then stack modifiers that capture long-tail variations.

A practical assembly method: pick one head term, then add three to five modifier tokens that each represent a distinct real query — material, use case, target user, capacity, color. Each token you add is a potential long-tail match, but only if it is a phrase people actually type.

• Good shape: 브랜드명 스테인리스 텀블러 500ml 보온 보냉 휴대용 (brand + stainless tumbler + 500ml + hot/cold retention + portable)
• Bad shape: 최고급 프리미엄 정품 인기 상품 텀블러 (adjective spam first, head term buried last)
• Keep spec numbers in the format Koreans type: 500ml, 1+1, 2개입 (2-pack), 대용량 (large capacity).
• Include one English token if your category has bilingual searchers — Coupang matches Latin script literally.

Step 3: Spacing, particles, and the details MT gets wrong

Korean spacing (띄어쓰기, ttuieossugi) is a real ranking variable. "무선청소기" and "무선 청소기" are searched in both forms, and tokenization is not always forgiving. When both spellings have volume, the safest move is to include the compound once and the spaced form once across the title and attribute fields rather than repeating it awkwardly in the title.

Other MT artifacts to strip out:

• Grammatical particles (은/는/이/가/을/를) — titles are noun stacks, particles just consume characters and add nothing to matching.
• Verb endings like ~합니다 or ~해요. No Korean listing title uses polite sentence endings.
• Over-literal renderings: "kitchen helper" becoming 주방 조력자 instead of the actual searched term 주방용품 (kitchen goods).
• Romanized brand names when the Hangul transliteration is what people search — check both, include the one with demand, and often both.
• Symbols and emoji. Many are stripped or penalized by marketplace title rules, and they eat your visible character budget.

Step 4: Validate with data before you publish

Never ship a title based on a guess about which synonym wins. Pull search volume for every candidate term, compare monthly demand and competition, then keep the top one or two and discard the rest. This single step separates listings that get impressions from listings that sit at zero.

This is the workflow SeoulRank was built around — Korean keyword volumes and Coupang daily rank tracking presented in English, so you can compare 텀블러 against 물통 against 보온병 without needing to read Korean or trust a translator's instinct.

After publishing, treat the title as a testable asset. Track your rank for the three to five phrases you deliberately targeted. If a phrase does not move within a couple of weeks, that token is dead weight — swap it for the next candidate on your list.

• Log your target phrases before launch so you know what "working" looks like.
• Change one variable at a time; rewriting the whole title makes results unreadable.
• Watch competitor titles in your category — they encode a lot of tested keyword knowledge for free.
• Re-check seasonal terms quarterly; Korean shopping language shifts fast around trends and holidays.

A quick pre-publish checklist

Run every listing through this before it goes live. It takes two minutes and catches most of the damage.

• Does the highest-volume head keyword appear in the first 25 characters?
• Have you removed all particles, verb endings, and filler adjectives?
• Are 3–5 distinct long-tail modifiers present, each backed by real search volume?
• Are spec tokens written the way Korean buyers type them (500ml, 2개입, 1+1)?
• Have you checked loanword vs. native word volume rather than assuming?
• Does the title still read as a natural Korean listing, not a keyword dump? Coupang and buyers both punish stuffing.