On this page
Why gates, not just "review"Seven gates for a Korean launchA RACI matrix for sign-offWhat each gate catchesFitting gates into a fast scheduleLaunch day. The Korean press release went out at nine. By eleven, a screenshot of the Korean landing page is circulating with a price that does not match the press release, because marketing changed the English price on Tuesday and nobody re-sent the page. The translation was fine. The process had no gate that would have caught it.
Public backlash over translation is rarely about a translator's skill. It is about the path text takes from source to publication, and who is supposed to stop it when something is wrong. This post sets out a sequence of review gates for a Korean launch, and a RACI matrix so every gate has an owner.
Why gates, not just "review"
"Have someone Korean look at it" is a review. A gate is a point the content cannot pass without a named person saying yes. The difference matters under deadline pressure, because reviews get shortened and gates get noticed when they are skipped.
The translation steps themselves are well defined. ISO 17100, the translation services standard, sets out the translation process as translation, check, revision, review, proofreading, and final verification and release. Those steps cover the text. A launch also needs gates around the text: whether the source was final, whether the Korean works on screen, and whether regulated claims were approved.
Seven gates for a Korean launch
- Source freeze. Content owners confirm the English is final for this release, and late changes go through a tracked change list. Strings are externalised and have context notes and screenshots.
- Translation and revision. Translation with self-check, then revision by a second linguist against the source and the termbase.
- In-context review. A Korean reviewer checks the build, the rendered page or the printed proof: truncation, line breaks, particles after variables, speech level, images with English text.
- Regulated content sign-off. Labels, terms, privacy text, pricing and claims checked by legal or regulatory owners against Korean requirements.
- Market review. The Korean market owner reads launch messaging as a customer would: tone, cultural fit, consistency with local channels.
- Final verification and release. Numbers, dates, prices and links re-checked against the final source and against each other across channels.
- Post-launch watch. For the first days, someone reads Korean reviews, social mentions and support tickets, with authority to fix copy quickly.
A RACI matrix for sign-off
RACI assigns four roles per task. PMI's description: Responsible is the person who does the work, and a task can have more than one; Accountable is the one person who owns the outcome; Consulted are people whose input is sought before the work is completed; Informed are kept up to date. The rule worth enforcing is that every task has exactly one Accountable person.
| Gate | Product / marketing owner | Localization manager | Translator / reviser | Korean market reviewer | Legal / regulatory | Engineering / QA |
|---|---|---|---|---|---|---|
| 1. Source freeze | A | R | I | I | C | R |
| 2. Translation and revision | I | A | R | C | I | I |
| 3. In-context review | I | A | C | R | I | R |
| 4. Regulated content | C | R | C | I | A | I |
| 5. Market review | A | C | C | R | I | I |
| 6. Final verification | A | R | I | C | C | R |
| 7. Post-launch watch | A | R | C | R | I | C |
This is an illustrative allocation. Small teams will merge roles, which is fine as long as each row still has one A. The common failure patterns are easy to spot in a matrix like this: a row with no A, where everybody assumed somebody else had approved; a row with two A's, where each owner waited for the other; and a Korean reviewer who is only ever C, consulted but never able to stop a release.
What each gate catches
| Gate | Typical error it stops |
|---|---|
| Source freeze | Korean translated from an outdated English draft |
| Translation and revision | Mistranslation, omission, inconsistent terms |
| In-context review | Truncated buttons, broken particles, English left in images |
| Regulated content | Claims that are fine in English but not permitted in Korean advertising or labeling |
| Market review | Tone or imagery that reads badly in Korea |
| Final verification | Mismatched prices, dates and links between channels |
| Post-launch watch | Anything that got through, fixed in hours instead of weeks |
Fitting gates into a fast schedule
Gates do not have to slow a launch if they are planned from the start. Work backwards from launch day: final verification the day before, in-context review two or three days before, translation and revision before that, and a source freeze that leaves room for all of it. Content that changes late, such as prices and dates, can be held as variables and filled from one source of truth, so a change made once updates every language. We discuss the trade-offs in speed vs quality in simultaneous shipping and the automation side in continuous localization automation.
Not every piece of content needs all seven gates. A blog post may skip gate 4; a warning label may need gate 4 twice. Match gates to the quality level of the content, and use risk-based sampling for high-volume, low-risk text. For the kinds of small errors these gates catch, see translation mistakes that damage brand trust.
Launching in Korea soon?
We can act as your translation, in-context review and final verification gates, with a named reviewer who can say no. Plan the launch with us →