Running your practice
Localization QA: How Translators and Agencies Run LQA Before Release
Localization QA (LQA) is the pre-release pass where translators and reviewers score a localized build against a weighted MQM-style rubric before shipping.
Localization QA sits as the final production gate inside the end-to-end Application Localization: The Translator's and Agency's Guide workflow, where translators and reviewers evaluate a localized build against a weighted MQM-style rubric before the client receives the staging build for sign-off.
What is localization QA (LQA)?
Localization QA is the pre-release production pass where translators and reviewers evaluate a localized build against a weighted MQM-style rubric scored per 1,000 words, catching linguistic defects, functional defects, and in-context usability failures before client sign-off. The scoring unit is errors per 1,000 words of localized content. The reviewer who scores an LQA pass cannot be the translator who produced the segment.
LQA covers five rubric dimensions in one pass:
- Accuracy — mistranslation, omission, addition.
- Terminology — term-base adherence.
- Regulatory compliance — locked boilerplate and brand terms unchanged.
- Readability — grammar, register, locale conventions.
- Usability — in-context defects visible in the running build.
How is LQA different from general software QA?
LQA scores locale-specific defects — mistranslation, terminology drift, cultural fit, text expansion breakage — against a translation rubric, while general software QA validates feature behavior against the spec in the source locale. A functional tester confirms the save button saves. An LQA tester confirms the Korean label on that button fits the control, matches the term base, and reads naturally to a native user.
Where does LQA sit inside the localization workflow?
LQA runs after translation and reviewer edits, after an Xbench or Verifika pre-pass, and before the client receives the staging build for sign-off. It is the final gate inside the parent application localization workflow. The translator's workspace holds five artifacts feeding into that gate: resource bundle, translation memory, term base, style guide, and a staging or pseudo-loc build.
What common challenges do localization QA testers face, and how can they be overcome?
LQA testers navigate tight deadlines, ambiguous style guides, missing source context, and inconsistent bug-tracking handoffs. Pairing each friction point with a concrete mitigation reduces cycle time and rework.
- Tight deadlines → agree a per-1,000-word review budget at scoping, not at hand-off.
- Ambiguous style guides → pre-LQA kickoff call with the reviewer and the translator.
- Missing source context → require screenshots or a staging build; refuse string-only review.
- Inconsistent bug records → enforce one record shape across linguistic and functional passes.
What is the difference between linguistic LQA and functional LQA?
Linguistic LQA checks translation accuracy, terminology, register, and cultural fit against the source. Functional LQA checks text truncation, broken placeholders, encoding errors, overlapping UI, and discoverability of translated features in the staging build. Both passes feed one bug tracker.
| Dimension | Linguistic LQA | Functional LQA |
|---|---|---|
| Primary input | Bilingual segments against source | Running staging build in target locale |
| Catches | Mistranslation, term drift, register, locale conventions | Truncation, placeholder breakage, mojibake, overlap, discoverability |
| Tester profile | Native-level linguist with MQM literacy | Native-level tester with build access and repro discipline |
| Primary tool | Xbench or Verifika, then TMS editor | Device/emulator, screenshot capture, Jira |
| Scoring | Weighted MQM categories per 1,000 words | MQM usability + design categories per 1,000 words |
What does the linguistic LQA pass catch?
Linguistic LQA catches mistranslation, omission, addition, term-base drift, wrong register, locale-convention breaks, and failures to localize culturally loaded content. It covers date formats, number formats, currency symbols, and address order. It scores under accuracy, terminology, and readability dimensions of the MQM rubric.
What does the functional LQA pass catch?
Functional LQA catches truncated strings, overlapping UI, broken placeholder tokens, encoding errors (mojibake), empty targets in the running build, and unreachable translated features. Each defect is logged with a screenshot and reproduction steps under the usability dimension.
How do in-UI discoverability gaps surface in functional LQA?
In-UI discoverability gaps surface when a feature ships fully translated but is only reachable via an unlabeled tab or icon-only control. LQA logs this as a usability error, not a translation error, and routes the UX fix back to the client. For the canonical write-up of how release-time LQA classifies these gaps, see Product Localization: How Translators and Agencies Deliver It.
What rubric and thresholds does an LQA pass use?
Production LQA uses a weighted MQM-style rubric scoring errors per 1,000 words across accuracy, terminology, regulatory compliance, readability, and usability. Critical content passes under 2 errors per 1,000 words; standard content passes under 5. Regulatory boilerplate and locked brand terms act as pass/fail blockers regardless of the per-1,000-word score.
What are the MQM error categories used in LQA?
MQM (Multidimensional Quality Metrics) is the industry-standard framework for categorizing and scoring LQA errors by type and severity. The categories used in LQA are:
- Accuracy — mistranslation, omission, addition.
- Terminology — term-base adherence, forbidden-term use.
- Linguistic conventions — grammar, punctuation, register, spelling.
- Style — compliance with the style guide.
- Locale conventions — dates, numbers, currency, address order.
- Design — truncation, overlap, line breaks.
- Verity — regulatory and factual correctness.
How is error severity scored (A/B/C) in an LQA bug record?
LQA bug severity runs on three levels:
| Severity | Definition | Release impact |
|---|---|---|
| A — Critical | Blocks release; regulatory or safety impact; locked brand term broken | Rejects the build regardless of error count |
| B — Major | Visible defect a user would notice; wrong meaning retained | Counts against the per-1,000-word threshold |
| C — Minor | Cosmetic or stylistic; preference-level | Counts at a lower weight |
Each record carries category, screenshot, repro steps, and a suggested fix in Jira or the studio tracker.
What release threshold separates a pass from a reject?
The release threshold used across our production work is under 2 errors per 1,000 words for critical content against the weighted MQM-style rubric. Any A-severity regulatory or locked-brand-term defect blocks release outright regardless of the per-1,000-word score. For how threshold failures manifest as shipped bad localization, see Bad Localization: What Goes Wrong and How to Avoid It.
Which tools do LQA testers actually use?
LQA testers run Xbench or Verifika as the pre-LQA linguistic pass, log defects in Jira or a studio tracker, validate i18n readiness with a pseudo-localization build, and use TMS-side QA profiles to catch number mismatches, placeholder breakage, and empty targets before the build is cut.
| Tool | Stage | Role in LQA |
|---|---|---|
| Xbench | Pre-LQA | Catches inconsistencies, number mismatches, empty targets, tag damage |
| Verifika | Pre-LQA | Alternative linguistic QA checker used before client handoff |
| TMS QA profiles (memoQ, Phrase, Trados) | In-translation | Blocks save on placeholder breakage, number mismatch, empty target |
| Pseudo-localization build | i18n gate | Verifies code accepts any locale before LQA opens |
| Jira (or studio tracker) | LQA pass | Default bug tracker across linguistic and functional defects |
| Visual regression / screenshot diff | LQA pass | Catches layout breakage against the pseudo-loc baseline |
What does Xbench or Verifika catch before LQA opens?
Xbench and Verifika catch inconsistencies, number mismatches, empty targets, double spaces, tag damage, and term-base violations before the build is cut. Running them removes static-file noise so human LQA time is spent on in-context defects, not errors a checker would surface in seconds.
How does pseudo-localization relate to LQA?
Pseudo-localization is the i18n gate that runs before LQA opens, verifying the code accepts any locale and surfacing truncation and encoding risks early. A build that fails pseudo-loc never reaches an LQA tester, protecting the per-1,000-word review budget. For the full mechanics, see Pseudo Localization: What Translators Need to Know About i18n Test Builds.
What automation exists beyond Selenium for LQA?
Beyond Selenium, LQA automation leans on TMS-side QA profiles, standalone linguistic checkers, screenshot diff tooling against pseudo-loc baselines, and locale-aware visual regression suites. Manual in-context play remains irreducible for cultural and discoverability defects. The automation layer narrows the surface the human tester covers; it does not replace the human.
What does an LQA bug record look like in production?
A production LQA bug record carries severity (A/B/C), category (linguistic, cosmetic, functional), locale, source string, target string, screenshot of the in-context defect, reproduction steps, device or build identifier, and a suggested fix. It is logged in Jira or the client's studio tracker.
| Field | Example value |
|---|---|
| Severity | B — major |
| Category | Functional — truncation |
| Locale | de-DE |
| Source string | Save changes |
| Target string | Änderungen speichern |
| Screenshot | Attached — settings panel, primary CTA |
| Repro steps | Open Settings → Account → scroll to footer |
| Device / build | iPhone 15, iOS 17.4, build 4721 |
| Suggested fix | Shorten to "Speichern" or widen button to 160px |
Pricing
How do agencies price LQA, and what do LQA testers earn?
US agencies price LQA scoring at $40–$90 per 1,000 words, alongside translation at $0.10–$0.25 per source word and review at $0.04–$0.08 per word. LQA testers earn hourly in a wide band depending on specialization, locale rarity, and whether work is onsite or remote.
How does LQA fit into an agency's three-band production quote?
LQA sits as the third band in a standard agency quote:
| Production band | US rate | Unit |
|---|---|---|
| Translation | $0.10–$0.25 | per source word |
| Review | $0.04–$0.08 | per word |
| LQA scoring | $40–$90 | per 1,000 words |
A mid-size program of 5 languages and 50,000 words runs $25,000–$75,000 end to end. LQA scoping fits into the project manager's dispatch; see Localization Management: How Translators and Agencies Run Multi-Locale Programs for how the LPM budgets pre-delivery QA on a single job.
What do LQA testers earn, and is the work full-time?
LQA testers are predominantly contract or project-based rather than full-time. AAA game studios recruit onsite for infosec reasons; Japan-market LQA roles almost always run as contract engagements, and native-level target-language skill is a hard requirement, not a preference. For how this plays out inside game studios, see Video Game Localization: What Translators and Agencies Need to Know.
How much does an LQA tester earn per hour?
Hourly rates for localization QA testers vary by language pair, market, and engagement type. Freelance rates typically range from $18–$45/hr, while in-house roles skew toward salaried bands. Gaming and regulated-industry work pays above that band; generalist SaaS LQA pays at or below it. Understanding hourly benchmarks helps hiring managers budget and testers negotiate fair compensation.
Get a quoteHow do you run LQA at a startup with no dedicated tester?
A minimum-viable LQA process at a startup runs an Xbench pre-pass on the exported bilingual files, a screenshot-based in-context review of the highest-risk screens, and one in-market reviewer with authority to block release — scored against a trimmed MQM rubric at the same per-1,000-word threshold.
- Export bilingual files and run Xbench against glossary and TM.
- Capture screenshots of the top 20 user-facing screens in the target locale.
- Hand to one in-market reviewer with explicit block-release authority.
- Log defects in Jira using the A/B/C severity and MQM category.
- Gate release at under 2 errors per 1,000 words for critical content.
Platform-constrained setups follow the same pattern inside a thinner tooling stack — see How Shopify Localization Works and What It Means for Your Agency.
Do screenshots count as a valid LQA method?
Screenshot review counts as a recognized LQA method for early-stage localization or startup budgets, provided screenshots are tied to source segments, carry locale and build identifiers, and feed the same bug record format a tester would produce in a staging build. Screenshots without source linkage are not evidence.
What qualifications does a localization QA tester need?
A localization QA tester needs native-level proficiency in the target language, working knowledge of a bug tracker (Jira), familiarity with MQM categories and severity scoring, and comfort inside TMS and QA checkers such as Xbench or Verifika. Entry roles in Japan typically require JLPT N1 plus a company-specific test.
Does an LQA tester need to be a native speaker?
LQA testers are hired for native-level target-language skill, not bilingual or advanced L2 ability. Game localization QA is especially strict: studios recruit for native fluency plus lived cultural reference in the target locale. Difficulty varies by locale — see Chinese Localization: What Makes It Harder Than Most Projects for one example of how writing-system scope raises the native-speaker bar.
Is LQA a stepping stone to translation, engineering, or localization management?
LQA is a recognized entry path into translation, localization project management, and software engineering QA. Hiring managers treat LQA experience as a net positive on engineering-adjacent resumes because it demonstrates bug-tracker discipline and release-gate judgment under production pressure.
About
What questions about localization QA come up most often?
The most common LQA questions concern the linguistic-vs-functional split, the pass threshold, screenshot validity, MQM severity disputes, and remote vs onsite work. Each has a short production answer.
Is LQA work typically remote or onsite?
LQA for AAA game studios is overwhelmingly onsite due to infosec restrictions on unreleased builds. SaaS and app LQA is more frequently remote, run against a staging build the agency's testers access through a time-boxed VPN or sandbox.
How do you dispute a flagged LQA error?
Error disputes are raised against the MQM category and severity assigned, with the translator citing source ambiguity, term-base precedent, or style-guide authority. The reviewer who did not translate the segment adjudicates. The final call is logged against the bug record and closes the loop.
How much does an LQA tester get paid?
LQA tester pay in the US tracks the agency scoring rate of $40–$90 per 1,000 words translated into hourly contract rates. Gaming and regulated-industry work pays above the band. Generalist SaaS LQA pays at or below it.
What is the difference between a localization QA tester and a localization tester?
The titles are often used interchangeably, but they describe different scopes. A localization tester focuses on functional and cosmetic rendering across locales; a localization QA tester owns the full quality gate — linguistic accuracy, MQM scoring, and release sign-off authority. On smaller teams the two roles collapse into one person. On larger programs they split, with the QA tester chairing the pass and the functional tester supplying device coverage. For how localized pages pass the same rubric at the SEO layer, see Localization SEO: How Translators Make Localized Pages Rank.
Frequently asked questions
What is a localization QA tester?
A localization QA tester is a bilingual reviewer who checks translated software, games, or documents for linguistic accuracy, cultural appropriateness, and functional correctness before release. Their work spans two overlapping disciplines: linguistic QA (checking translation quality against the source) and functional QA (verifying that the localized build actually works — no truncated strings, broken UI, or encoding errors). In translation agency contexts, the role may also include checking certified translations against style guides, glossaries, and regulatory formatting requirements.
What does a localization job do?
A localization professional adapts content — software, legal documents, marketing copy, certified translations — so it is linguistically accurate and culturally appropriate for a target market. The work spans translation, editing, terminology management, desktop publishing, and QA. In an agency setting this often means managing a full TEP (translation, editing, proofreading) workflow, coordinating with subject-matter reviewers, and ensuring final deliverables meet client or regulatory standards before sign-off.
What's the difference between linguistic QA and functional QA in localization?
Linguistic QA focuses on the quality of the translated text itself — accuracy, fluency, terminology consistency, and adherence to style guides — while functional QA tests whether the localized product works correctly as a piece of software or a formatted document. Linguistic QA is done by bilingual reviewers comparing source and target strings; functional QA is done by testers (sometimes monolingual) who click through a build, check that UI elements render correctly, strings are not truncated, date/number formats are right, and hyperlinks work. Both layers are required for a production-ready release.
What does an MQM error category and severity framework look like in practice?
MQM (Multidimensional Quality Metrics) organizes translation errors into a hierarchical taxonomy — top-level categories include Accuracy, Fluency, Terminology, Style, Locale Convention, and Verity — each subdivided into specific issue types. Each issue is also tagged with a severity: Minor, Major, or Critical. In practice, a reviewer annotates a segment like 'wrong term used for a regulated concept' as Terminology > Incorrect Term > Major, then scores the text by weighting errors against word count. A common passing threshold is fewer than 1–3 major errors per 1,000 words, though agencies set their own thresholds per project type.
Which tools can be used for automation of localization testing beyond Selenium?
Beyond Selenium, common tools for automated localization testing include Appium (mobile apps), Playwright and Cypress (modern web UIs), and pseudo-localization utilities built into platforms like Phrase or Lokalise. For string-level checks, QA tools such as Xbench, Verifika, and the built-in QA modules in memoQ or SDL Trados catch tag errors, terminology mismatches, and number inconsistencies automatically. CI/CD pipelines increasingly integrate linters that flag untranslated strings or encoding issues on every build commit.
Do I need N1 Japanese (or equivalent native fluency) to get an entry-level LQA job?
Native or near-native fluency in the target language is typically required for linguistic LQA roles, but formal certification like JLPT N1 is rarely mandated — demonstrated fluency and translation experience usually matter more to hiring managers. Functional LQA roles (testing UI rendering, layout, and functionality) can sometimes be filled by testers with strong target-language reading skills rather than full native fluency. For high-stakes domains like legal, medical, or certified translation review, native-level command is effectively non-negotiable.
Is LQA work typically full-time or contract/freelance?
LQA work skews heavily toward contract and freelance arrangements, especially for language-specific roles. Large game studios and global enterprises sometimes maintain small in-house LQA teams for their primary markets, but most agencies staff surge projects through freelance reviewers hired per-project or on short-term contracts. Rates for freelance linguistic QA typically range from $0.02–$0.06 per word or $25–$60 per hour depending on language pair and domain. Full-time in-house positions do exist, particularly for high-volume languages like Spanish, French, German, and Japanese.
How do I actually do LQA at a startup with limited resources — do screenshots count?
At resource-constrained startups, a pragmatic LQA process combines in-context screenshots with a structured linguistic review checklist — screenshots absolutely count as long as reviewers can see translated text in its rendered context. Prioritize a bilingual review of highest-visibility strings (UI labels, error messages, onboarding copy) using a lightweight error log based on MQM severity categories. Free or low-cost tools like Xbench, Google Sheets error trackers, and pseudo-localization passes can replace expensive QA platforms. Even a single native-speaker reviewer catching Critical errors before release adds significant quality value.
How do I break into game localization QA with no experience?
Start by building demonstrable language skills and a portfolio of mock LQA reports on publicly available game demos or open-source software localizations. Many game studios accept applications for junior LQA roles with no professional experience if you can pass a language and bug-reporting test — strong gaming knowledge in the target culture is treated as an asset. Connecting with localization communities (LocWorld, the Localization Institute, LinkedIn groups) and targeting mid-size studios or localization vendors rather than major publishers improves entry odds. A short course in CAT tools or JIRA/bug-tracking basics strengthens applications.
What is a localization job?
A localization job is a professional role focused on adapting content for a specific target language, culture, and market — going beyond word-for-word translation to ensure the product feels native to its audience. Roles range from translator and editor to project manager, LQA tester, desktop publishing specialist, and terminology manager. In the certified translation industry specifically, localization work also involves meeting regulatory formatting requirements — correct layouts, stamps, certifier statements — so that documents are accepted by government bodies, courts, or academic institutions in the target country.
Related
- Localization SEO: How Translators Make Localized Pages Rank
- Localization SaaS: How Translators and Agencies Run Continuous Localization for Software Clients
- Localization Solution: What Translators and Agencies Assemble to Run Multi-Lane Work
- Localization Workflow: How Translators and Agencies Wire the Production Pipeline
- Localization in Marketing: What Translators and Agencies Deliver
- Memory Translation Software: What It Does Inside a Translator's Workflow