Running your practice
Localization Solution: What Translators and Agencies Assemble to Run Multi-Lane Work
A localization solution is the stack translators and agencies assemble — TMS, CAT, MT engine, QA gates, connectors — to deliver multi-lane localized output per
Production-side localization is the stack a translator or agency assembles — not a single tool a client buys off a shelf. The reader here is the operator wiring TMS, CAT, MT, memory, QA, connectors, and reviewers into one program that serves commercial and certified lanes at once. For the parent context, see How App Localization Works and What Clients Are Buying.
What is a localization solution from the production side of the desk?
A localization solution is the assembled production stack — TMS, CAT tool, MT engine, translation memory, QA and LQA gates, connectors, and reviewer workflow — that translators and agencies compose per locale to deliver software, document, web, support, and certified lanes under one program. The solution is the factory. The services are what the factory ships.
Seven elements sit inside every production-side solution:
- TMS for string, file, and project management.
- CAT tool as the translator workbench.
- MT engine for first-draft output.
- Translation memory for reuse across lanes and locales.
- QA automation plus an LQA scorecard.
- Connectors to source systems, CMS, design, and support tools.
- Reviewer and certifier workflow on top.
How is a localization solution different from a translation management system?
A translation management system is one layer inside a localization solution. The solution adds the CAT tool, MT engine, TM, QA/LQA gates, connectors, and reviewer workflow that turn string management into deliverable, certifiable output. A TMS alone handles files and tasks. It does not certify a page, run an LQA scorecard, or produce a named-translator attestation.
Who assembles the solution — the client, the agency, or the TMS vendor?
On production-side work, the agency assembles the solution. The client supplies source files and locale targets. The TMS vendor supplies one layer. The agency wires the CAT, MT, TM, QA, connectors, and reviewer pool across every lane the contract covers. See Localization Management: How Translators and Agencies Run Multi-Locale Programs for the operating cadence.
What are the core functions a localization solution has to perform?
A localization solution performs seven core functions: string and file intake, translation memory leverage, machine translation pre-translate, human translation and review, QA automation, LQA scoring, and release connectors back to the client's build or CMS — with a certified-lane overlay when regulated content is in scope. Each function is a discrete stage that produces a defined artifact; removing any one stage breaks the chain and creates risk downstream. The full stack is only as strong as its weakest gate, which is why production-side operators treat every layer as mandatory rather than optional.
| Function | What it produces | Who owns it |
|---|---|---|
| Intake | Normalized strings and files | TMS layer |
| TM leverage | Reused approved segments | TM store |
| MT pre-translate | First-draft segments | MT engine |
| Translation + review | Target segments | Translator + reviewer |
| QA automation | Tag, number, placeholder checks | CAT/TMS |
| LQA | MQM-style scorecard output | LQA reviewer |
| Release connector | Approved strings back to build/CMS | Connector layer |
How does continuous localization fit into the solution?
Continuous localization is the operational mode that turns a localization solution from a batch-file service into a live production lane. Connectors poll the client's source repository on every commit, push new or changed strings into the TMS automatically, assign them to the appropriate translators and reviewers, and return approved target strings directly to the build — eliminating the manual export/import cycle that creates version lag and error risk. The connector layer is what allows a multi-locale program to keep pace with weekly or daily product releases without a separate handoff step. Platforms purpose-built for continuous localization expose this mechanism through API-native integrations with Git hosts, CI/CD pipelines, and headless CMS environments, so that translation management improves turn-around speed at the infrastructure level rather than by asking human operators to work faster.
Where does machine translation sit as a stage, not an answer?
Machine translation is a pre-translate stage inside the solution, not the solution itself. An MT engine drafts segments and deposits them into the CAT workbench. The TM layer overrides any segment it has an approved match for, because a human-validated prior translation always outranks a fresh MT draft. A post-editor then works through remaining MT output segment by segment before the content advances to the review gate. No MT output is ever delivered as finished certified translation without a qualified human reviewer signing off — a constraint that holds across regulated, legal, and government-filing lanes regardless of MT quality. Machine translation and professional translation services are complementary layers in the stack, not substitutes for each other.
What does translation memory and terminology management contribute?
Translation memory stores every approved segment pair for reuse across lanes and locales, making it the compounding asset of the entire solution. Every project that runs through the stack adds to TM depth, which in turn raises leverage rates on future projects and lowers per-word costs for the client over time. Terminology management enforces one term base per client so brand voice, product names, and regulated phrasing stay consistent across all seven lanes. One TM lives per client per locale, carrying lane-level metadata so that a certified-document segment does not overwrite a marketing variant of the same source string — a critical separation that protects tone, register, and legal accuracy simultaneously.
What are the seven layers of a production-side localization solution?
A production-side localization solution has seven layers: TMS, CAT tool, MT engine, translation memory, QA automation plus LQA scorecard, connectors, and reviewer and certifier workflow — each with its own vendor shortlist and pricing line.
| Layer | Function | Typical choices |
|---|---|---|
| 1. TMS | Strings, files, projects | Crowdin, Phrase, Lokalise, Lingo.dev |
| 2. CAT tool | Translator workbench, audit trail | memoQ, SDL Trados |
| 3. MT engine | First-draft output | DeepL, Google, Amazon, Microsoft, LLMs |
| 4. TM | Segment reuse | Native to TMS; one per client per locale |
| 5. QA + LQA | Automated checks and scored samples | CAT/TMS QA; MQM scorecard |
| 6. Connectors | Source, CMS, design, support, build | GitHub, Contentful, Figma, Zendesk, CI |
| 7. Reviewer/certifier | Sign-off and attestation | Named translator + LQA reviewer |
Which TMS platforms do agencies actually standardize on?
Agencies serving SaaS and multi-lane clients standardize on Crowdin, Phrase, Lokalise, or Lingo.dev. Crowdin fits open-source and mid-market SaaS with GitHub integration. Phrase fits enterprise TM and QA depth. Lokalise fits mobile-first products. Lingo.dev and similar Git-native tools fit developer-led teams.
Which CAT tools pair with the TMS for certified and high-stakes content?
memoQ and SDL Trados are the CAT tools agencies pair with their TMS when certified translation, legal, or regulated content enters the stack. They carry the segment-level audit trail, bilingual review views, and offline working mode that a certified lane and an ISO 17100-aligned workflow demand.
Which MT engines plug into the solution and how are they chosen?
Agencies plug DeepL, Google Translate, Amazon Translate, Microsoft Translator, and LLM-based engines (GPT, Claude) into the TMS as swappable MT layers. Choice is made per locale and per lane:
- BLEU and COMET scores for the language pair.
- Domain fit (legal, medical, UI, marketing).
- Post-edit distance measured on recent jobs.
- Regulated lanes default to the lowest post-edit effort engine for that pair.
What QA and LQA gates run before client delivery?
Two gates run before client delivery. Automated QA inside the CAT/TMS checks tags, numbers, terminology, and placeholders on every segment. LQA samples output at $40–$90 per 1,000 words against an MQM-style scorecard aligned to ISO 17100. The five scorecard categories are accuracy, fluency, terminology, locale convention, and style. Certified lanes carry a mandatory sign-off on the certificate of accuracy.
What connectors does the solution need to earn its 'continuous' label?
A continuous solution needs connectors to the client's source repo, CMS, design tool, help center, and build pipeline:
- Source: GitHub, GitLab, Bitbucket.
- CMS: Contentful, Sanity, WordPress, Drupal.
- Design: Figma.
- Support: Zendesk, Intercom.
- Build: CI webhooks.
Framework-level hooks are covered in React i18n: How Translators Fit Into a React Localization Pipeline. For locale directory and hreflang coverage downstream, see Localization SEO: How Translators Make Localized Pages Rank.
How does the reviewer and certifier workflow sit on top?
The reviewer and certifier workflow is the seventh layer. Credentialed reviewers are assigned per lane rather than per file. Certified-lane work routes to a named translator who signs the certificate of accuracy on every page. Notarization is added as an optional $10–$50 line when the receiving authority requires it.
How does team collaboration work inside a production-side localization solution?
A solution handling certified and high-stakes content supports structured collaboration — task assignment, ownership tracking, and bottleneck surfacing across translators, reviewers, and certifiers. Three collaboration patterns run concurrently:
- Round-robin assignment inside the TMS for commercial-lane strings.
- Named-translator routing in the CAT tool for certified pages.
- LQA reviewer queue sampling across both streams against the scorecard.
Which content lanes does one localization solution have to serve at once?
One localization solution typically serves seven content lanes concurrently: product/app strings, website, marketing, support knowledge base, legal and compliance, in-product video, and certified translation. All seven share one term base and one style guide per client per locale, with lane-specific workflow templates and SLAs layered on top. The simultaneous operation of these lanes is not an edge case — it is the default operating condition for any agency running a mid-market or enterprise program. Lane isolation at the workflow level is what prevents a fast-moving marketing cycle from contaminating the audit trail required by the legal or certified lane.
Lane coverage inside one program:
- Product/app strings — UI and in-app copy routed through Git-connected TMS workflows with automated QA on tags and placeholders.
- Website — page-level and component-level content managed through CMS connectors, with hreflang metadata emitted at release.
- Marketing collateral and campaigns — transcreation-heavy content requiring locale-specific style rules and separate reviewer pools.
- Support knowledge base articles — high-volume, fast-turnaround content where TM leverage and MT post-edit deliver the strongest cost offset.
- Legal and compliance text — reviewed against jurisdiction-specific term bases; flagged for mandatory human sign-off before release.
- In-product video — subtitle files and voiceover scripts handled as time-coded segments with character and reading-speed constraints.
- Certified translation — regulated filing lane with locked boilerplate, named-translator attestation, and mandatory LQA before delivery.
Where does certified translation production sit inside the solution?
Certified translation is the seventh regulated lane, sitting alongside product/app, website, marketing, support KB, legal/compliance, and in-product video. It is not a separate product delivered outside the solution stack — it is a lane overlay that adds specific workflow controls on top of the shared infrastructure. The lane requires a competent-translator certificate, shares the client's term base and style guide with the other six lanes, and adds locked boilerplate for the certificate of accuracy plus a named-translator attestation on every page. Pricing for the certified lane runs $0.10–$0.25 per source word or
One localization solution typically serves seven content lanes concurrently: product/app strings, website, marketing, support knowledge base, legal and compliance, in-product video, and certified translation. All seven share one term base and one style guide per client per locale, with lane-specific workflow templates and SLAs on top. See Service Localization: What Translators and Agencies Actually Deliver for the lane-by-lane deliverables.
Lane coverage inside one program:
- Product/app strings — detailed in Product Localization: How Translators and Agencies Deliver It.
- Website — detailed in Website Localization: What Translators and Agencies Actually Do on the Project.
- Marketing collateral and campaigns.
- Support knowledge base articles.
- Legal and compliance text.
- In-product video (subtitles, voiceover scripts).
- Certified translation for regulated filings.
Where does certified translation production sit inside the solution?
Certified translation is a seventh regulated lane sitting alongside product app, website, marketing, support KB, legal/compliance, and in-product video. It carries a competent-translator certificate requirement. It shares the term base and style guide with the other six. It adds locked boilerplate for the certificate of accuracy plus a named-translator attestation on every page. The lane prices at $0.10–$0.25 per source word or $25–$150 per standard page, with notarization at $10–$50 when the receiving authority requests it.
How do software, document, and web lanes share the same TM?
Software, document, and web lanes share the same TM by writing every approved segment back to one TM per client per locale, with lane-level metadata. Lane metadata keeps a certified-document segment from overwriting a marketing variant. Leverage reports can then be attributed to the lane that earned them.
5–
What does a localization solution cost in the US — line by line?
In the US, a localization solution prices in five production lines plus tooling:
| Line | Unit | US range |
|---|---|---|
| Translation | per source word | $0.10–$0.25 |
| Review | per source word | $0.04–$0.08 |
| LQA | per 1,000 words | $40–$90 |
| Project management | % of project value | 10–15% |
| Certified page | per standard page | $25–$150 |
| Notarization (optional) | per document | $10–$50 |
What does a mid-size 5-locale, 50,000-word program budget to?
A mid-size program covering 5 languages and 50,000 words budgets at $25,000–$75,000 end to end. Four drivers push the landing point inside that band:
- TM leverage (how much repeat content exists).
- Turnaround speed (rush multipliers).
- Certification requirements (named-translator overhead).
- Whether an AI pre-translate stage shifts the cost mix.
For the corporate-client framing of the same spend, see Business Localization: How Agencies Deliver Multi-Locale Programs for Corporate Clients.
How does AI pre-translate reshape the cost mix?
AI pre-translate compresses the translation line toward post-edit rates and pushes more of the quality burden onto review and LQA. Review stays at $0.04–$0.08 per word. LQA stays at $40–$90 per 1,000 words. The published cost base holds. The ratio between the five lines shifts away from raw translation and toward QA sign-off.
Get a quoteHow do agencies choose the right localization solution for the book of business?
Agencies choose by mapping four selection criteria against the book of business: lane mix, locale count and tier, regulated-content share, and developer-versus-content-led client type. The shortlist is then any TMS + CAT + MT combination that satisfies all four without requiring the agency to maintain duplicate infrastructure for different client profiles. A solution that handles only developer-led, Git-native clients will break down the moment a regulated-filing lane enters the scope, because that lane demands a segment-level audit trail and named-translator attestation that a lightweight TMS does not produce by default.
| Client profile | Primary TMS fit | CAT overlay |
|---|---|---|
| Mid-market SaaS, Git-centric | Git-connected TMS with CI/CD hooks | Desktop CAT for certified work |
| Enterprise, deep TM and QA | Full-featured TMS with LQA gate | Legal/regulated CAT workbench |
| Mobile-first product | App-native TMS with OTA support | Desktop CAT for document lanes |
| Developer-led, API-first | API-native TMS, Git webhook triggers | Desktop CAT on demand |
What ISO standards does the solution have to respect?
A production-side localization solution is required to respect three ISO families: ISO 17100 for translation service quality and LQA scorecard structure, ISO 639 for language codes, and ISO 3166-1 for country subtags. ISO 17100 is the anchor standard for the certified lane — it defines translator competence requirements, the review step, and the documentation that constitutes a complete quality record. ISO 639-1 holds the two-letter codes used in most software i18n frameworks. ISO 639-2 extends coverage to approximately 490 languages with three-letter codes. ISO 639-3 covers roughly 7,000 languages for agencies operating in less-resourced locale markets. Combined with ISO 3166-1 country subtags, these codes produce BCP 47 locale tags — es-MX, pt-BR, fr-CA — that the web lane embeds in hreflang pairs at release. The LQA scorecard aligned to ISO 17100 is priced at $40–$90 per 1,000 words translated and is mandatory on certified lanes; skipping it is not a cost-saving option — it is a compliance gap.
What does the solution need for the certified lane specifically?
The certified lane imposes the most demanding set of solution requirements because it produces a legal artifact rather than a user-facing string. The solution needs:
Optional notarization at
What does a localization solution look like in production — a worked example?
A concrete production example illustrates how all seven layers of the stack operate together inside a single program. Consider a US agency running a five-locale program for a SaaS client: the agency wires a Git-connected TMS to the client's source repository, pre-translates new strings with an MT engine, routes post-edit work to in-country translators in a desktop CAT workbench, runs automated QA plus LQA at $40–$90 per 1,000 words translated, and spins certified-document work out of the same translation memory into a named-translator attestation workflow for regulatory filings at
A worked example: a US agency running a SaaS client's 5-locale program wires Crowdin to the client's GitHub repo, pre-translates with DeepL, routes post-edit to in-country translators in memoQ, runs automated QA plus LQA at $40–$90 per 1,000 words, and spins certified-document work out of the same TM into a named-translator attestation workflow for USCIS filings at $25–$150 per page.
The sequence per commit:
- GitHub commit triggers the Crowdin connector.
- Changed strings land in the TMS; TM leverage runs.
- DeepL pre-translates unmatched segments.
- Translators post-edit in memoQ.
- Reviewers mark terminology and locale fit.
- Automated QA clears tags, numbers, placeholders.
- LQA samples segments against the MQM scorecard.
- Approved strings return to the build; certified pages route to the named translator.
5–
Frequently asked questions
What does "localization" mean?
Localization is the process of adapting content — text, images, formats, and cultural references — so it feels native to a target market, not just translated. It goes beyond word-for-word translation to cover date formats, currencies, units of measure, legal conventions, and tone. For certified translation production specifically, localization means ensuring that documents meet the linguistic and regulatory standards of the destination country, not just the destination language. A birth certificate translated for use in Brazil, for example, requires different formatting norms and certification language than the same document destined for Spain.
What does localize mean?
To localize something means to rework it so it reads, looks, and functions as if it were originally created for a specific locale. In a translation workflow, localizing a certified document involves adapting terminology to the target jurisdiction's legal register, adjusting name-order conventions, reformatting dates, and framing the translator's certification statement to comply with local requirements. It is distinct from simple translation: a localized document passes as native; a merely translated one often reads as foreign.
What does localization mean in software?
In software, localization (often abbreviated L10n) means adapting a product's interface, strings, date/time formats, and regional settings for a specific market after the internationalization (i18n) framework has been built. For translation agencies managing multi-lane workflows, software localization is one of the highest-volume project types: a single app release can involve 20–40 target languages simultaneously, requiring coordinated CAT-tool projects, glossary enforcement, and QA passes across all lanes. Certified translations of legal or compliance strings within that software add an additional layer of accountability.
What is another term for localization?
"Transcreation" and "cultural adaptation" are the terms most often used interchangeably with localization, though each has a slightly different emphasis. Transcreation stresses creative rewriting — common in marketing — while cultural adaptation is used in regulated fields like medical or legal translation. In the certified translation industry, practitioners may also use "target-market adaptation" or simply "regional compliance" when describing how a translated document is aligned with the legal and bureaucratic norms of the destination jurisdiction.
How does localization work?
Localization follows a structured pipeline: source content is analyzed, a glossary and style guide are created, translators produce drafts, editors review for linguistic and cultural accuracy, and a final QA pass checks formatting and compliance. In certified translation production, an additional step requires a qualified translator to sign or stamp a certification statement attesting to accuracy. Agencies running multi-lane work manage these steps in parallel across language pairs, using translation management systems (TMS) and CAT tools to track progress, enforce terminology, and deliver consistent output at scale.
How much does localization cost?
Localization costs typically range from $0.10 to $0.40 per source word for professional translation, with certified documents commanding a premium of $30–$150 per page depending on language pair and jurisdiction. Project management, desktop publishing, engineering, and QA can add 20–40% on top of the per-word rate. Rush delivery, rare language pairs, and notarization requirements push costs higher. Agencies using first-draft automation tools can reduce the per-word labor cost on repetitive certified document types, improving margins without compromising the required human review and certification step.
How much do translators charge per 1000 words?
Professional translators typically charge $80–$200 per 1,000 words depending on language pair, subject matter, and whether certification is required. Common European language pairs (e.g., Spanish–English) sit toward the lower end, while rare pairs or highly technical fields — legal, medical, patent — command rates of $150–$250 per 1,000 words or more. Certified translations often carry a flat per-page or per-document fee on top of the per-word rate. Translators working with first-draft automation can process higher volumes at these rates, directly improving hourly yield.
How much should I charge for translation per hour?
Most freelance translators bill $40–$100 per hour when working on an hourly basis, though per-word pricing is more common in the industry. Certified translation specialists — particularly in legal, medical, or immigration document fields — often earn toward the upper end of that range or beyond it. Hourly billing makes most sense for tasks like editing, revision, or consultation where output volume is unpredictable. Translators who use productivity tools and first-draft automation can effectively raise their hourly yield even when billing per word, since they process more words in the same time.
What are the five types of translation?
The five types most commonly recognized in the industry are: literary translation, technical translation, legal translation, medical/scientific translation, and audiovisual (or multimedia) translation. Certified translation — the focus of production workflows for agencies and professional translators — falls primarily under legal translation but applies across all types when a government body or institution requires an attested, signed declaration of accuracy. Each type demands different subject-matter expertise, terminology resources, and QA standards, which is why agencies often segment their translator pools by specialization.
What is an example of localization?
A concrete example in certified translation production is adapting an academic transcript for university admission in Germany: the translator not only converts the text from English to German but also maps U.S. grading scales (GPA) to German equivalents, reformats dates to DD.MM.YYYY, uses the appropriate German administrative register, and appends a certification statement worded to meet German authority requirements. Simply translating the words would not produce a document that German institutions accept — the cultural and regulatory adaptation is what makes it a true localization.
Related
- 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
- OmegaT for Professional Translators: What It Does and What It Lacks
- Pharma Localization: What Translators and Agencies Actually Deliver
- Phrase AI: What It Does for Translators and Where It Fits the Certified-Translation Stack