The international standard for systematic translation quality evaluation. Move beyond subjective assessments to objective, repeatable quality metrics. Standardize your QA process with internationally recognized methodology.
ISO 21999:2020, officially titled "Translation quality assurance and assessment -- Models and metrics," provides a comprehensive framework for evaluating translation quality through systematic, objective, and repeatable methods. It establishes models and metrics that enable translation service providers, clients, and evaluators to assess quality in a consistent and comparable way.
The translation industry has long struggled with a fundamental challenge: how do you objectively measure translation quality? Different reviewers, different clients, and different agencies have historically used different criteria, different error categories, and different scoring methods. What one reviewer considers a critical error, another might categorize as minor. What one client considers acceptable quality, another might reject. This lack of standardization leads to disputes, inconsistent expectations, and an inability to benchmark and improve quality over time.
ISO 21999 solves this problem by providing a standardized framework that defines how to categorize translation errors, how to assign severity levels, how to calculate quality scores, and how to make pass/fail decisions based on objective criteria. It transforms quality evaluation from a subjective art into a systematic science, while still acknowledging that translation quality ultimately serves human communication purposes.
Before ISO 21999, translation quality evaluation was often a source of friction between agencies and clients. Common problems included:
ISO 17100 and ISO 21999 serve different but complementary purposes in the translation quality ecosystem. ISO 17100 focuses on the translation production process: it defines the qualifications translators need, the workflow they must follow (translation, check, revision), and the management systems that support quality delivery. It answers the question: "How should translations be produced to ensure quality?"
ISO 21999 focuses on translation quality evaluation: it defines how to categorize errors, measure quality, and make objective quality assessments. It answers the question: "How do we measure whether translations meet quality requirements?" Together, these standards create a complete quality loop: ISO 17100 ensures quality is built into the production process, and ISO 21999 ensures quality is measured and verified after production.
This complementary relationship means that agencies certified to both standards can demonstrate not only that they follow rigorous production processes, but also that they have objective, documented methods for verifying the quality of their output. This dual certification provides the most comprehensive quality assurance story in the translation industry.
Transform quality evaluation from subjective opinion into objective, data-driven assessment.
Replace subjective "good enough" assessments with standardized, measurable quality scores. Track quality trends over time, compare performance across language pairs, and provide clients with data-driven quality reports that demonstrate your commitment to excellence.
When clients raise quality concerns, an ISO 21999-compliant evaluation provides an objective framework for discussion. Instead of subjective disagreements, both parties can reference standardized error categories, severity definitions, and quality thresholds to reach fair conclusions.
Systematic error categorization reveals patterns that enable targeted improvement. If most errors are terminology-related, you invest in better glossaries. If accuracy errors cluster in certain domains, you adjust translator assignment. Data drives improvement, not guesswork.
Standardized quality metrics provide a fair, transparent basis for evaluating translator and vendor performance. Instead of subjective assessments, you can track quality scores over time, identify top performers, and provide constructive, data-based feedback.
ISO 21999 quality frameworks can be implemented in modern CAT tools and TMS platforms, enabling automated quality scoring, real-time error tracking, and integrated quality management within your existing translation technology stack.
Demonstrating a certified, standards-based approach to quality evaluation gives clients confidence that quality is not left to chance. They know that every translation is evaluated against objective criteria, and that quality issues are systematically identified and addressed.
A systematic approach to identifying, categorizing, and scoring translation errors.
The most critical category. Includes mistranslation (incorrect meaning transfer), omission (missing content), addition (content not in source), and untranslated text. Accuracy errors directly affect whether the translation fulfills its communicative purpose. Sub-categories include factual errors, numerical errors, and reference errors that could have serious consequences.
Covers target language correctness: grammatical errors (agreement, tense, case), spelling and typographical errors, punctuation errors, and syntactic issues. While less critical than accuracy errors, linguistic errors affect readability and professionalism. The framework distinguishes between errors that affect comprehension and those that are purely cosmetic.
Addresses domain-specific vocabulary and client terminology compliance. Includes use of incorrect terms, inconsistent terminology within a document, deviation from approved glossaries, and failure to use client-preferred terminology. This category is especially important in technical, medical, and legal translation where precise terminology is essential.
Covers stylistic aspects: inappropriate register (too formal or informal), tone inconsistency, poor readability, awkward phrasing, and deviations from client style guides. Style errors may not affect accuracy but impact the user experience and brand consistency. The framework helps distinguish style preferences from genuine style errors.
Addresses locale-specific requirements: date and number formatting, measurement unit conversion, currency representation, cultural conventions, and regional language variants. These errors can range from minor formatting issues to serious cultural inappropriateness depending on context and severity.
Each error is assigned a severity level that affects the quality score. Critical: errors that could cause harm, legal issues, or complete misunderstanding. Major: errors that significantly affect meaning or usability. Minor: errors that are noticeable but do not significantly affect comprehension. Preferential: alternative renderings that are not errors per se but differ from evaluator preference.
One of the most valuable contributions of ISO 21999 is establishing a clear distinction between objective error-based evaluation and subjective quality assessment. Both have their place, but they serve different purposes and should not be conflated.
Analytical evaluation uses the error categorization and severity framework to produce quantitative quality scores. A defined sample of the translation is reviewed against the source text, errors are identified and categorized, severity levels are assigned, and a quality score is calculated using a weighted formula. This approach produces repeatable, comparable results that can be tracked over time and benchmarked against defined quality thresholds.
The analytical approach is ideal for: ongoing quality monitoring, translator performance evaluation, SLA compliance verification, quality trend analysis, and dispute resolution. It provides the hard data that drives informed decision-making about quality management.
Holistic evaluation assesses the overall quality of a translation as a whole, considering factors like fluency, naturalness, communicative effectiveness, and fitness for purpose. While more subjective, it captures aspects of quality that error counting alone may miss, such as overall readability, creative adaptation quality, and the degree to which the translation achieves its intended purpose.
The holistic approach is ideal for: creative content evaluation, marketing transcreation assessment, final quality sign-off, and client acceptance reviews. It complements analytical evaluation by providing the human judgment perspective that pure metrics cannot capture.
Modern translation technology platforms can implement ISO 21999's quality framework directly into the translation workflow. This includes configurable error taxonomies aligned with ISO 21999 categories, automated quality scoring based on reviewer annotations, dashboard reporting showing quality trends across projects, language pairs, and translators, integration with revision workflows so quality data feeds back into production processes, and API-based quality reporting for client portals and automated notifications.
Major CAT tools and TMS platforms including memoQ, Trados, Phrase, XTM, and Memsource support customizable QA frameworks that can be configured to align with ISO 21999 categories and severity levels. TranslationCert can guide you through configuring your existing tools to support your certified quality evaluation framework.
Implement a world-class quality evaluation framework with expert guidance at every step.
We evaluate your current quality evaluation practices, error categorization methods, quality scoring approaches, and QA tools. You receive a detailed analysis of how your current practices compare to ISO 21999 requirements, with specific recommendations for alignment.
1-2 daysWe help you design and document your quality evaluation framework aligned with ISO 21999. This includes error taxonomy definition, severity level criteria, quality scoring methodology, sampling procedures, evaluator guidelines, and integration plans for your CAT tools and TMS.
3-5 daysImplement the quality framework across your operations and calibrate your reviewers. This includes training evaluators on the error taxonomy, conducting inter-rater reliability exercises to ensure consistency, configuring QA tools, and running pilot evaluations to refine the framework.
1-2 weeksQualified auditors review your quality evaluation framework, examining documentation, sample evaluations, evaluator training records, quality scoring methodology, and evidence of framework implementation. The audit verifies that your approach meets ISO 21999 requirements.
1-2 daysUpon successful completion, your ISO 21999 certificate is issued. You can demonstrate your standards-based quality evaluation approach to clients, include it in proposals, and use it as a competitive differentiator. Valid for 3 years with annual surveillance audits.
2-3 daysElevate your quality assurance from subjective opinion to certified, objective evaluation.
Move from gut feeling to hard data. Track quality scores across projects, language pairs, and translators to identify trends, measure improvement, and make informed resource allocation decisions.
Standardized error categories and severity definitions provide an objective framework for discussing quality with clients, reducing subjective disagreements and building stronger relationships.
Error categorization data reveals where to focus improvement efforts. If terminology errors dominate, invest in glossaries. If accuracy drops in certain domains, adjust translator qualification requirements.
Objective quality scores provide a transparent, unbiased basis for evaluating and compensating translators, improving morale and retention among your top performers.
Together, ISO 17100 (production quality) and ISO 21999 (evaluation quality) create the most complete quality assurance story in the industry, demonstrating excellence in both delivery and measurement.
The framework integrates seamlessly with modern CAT tools and TMS platforms, enabling automated quality tracking without adding manual overhead to your existing workflows.
Everything you need to know about ISO 21999 translation quality evaluation certification.
ISO 17100 focuses on the translation production process -- it defines how translations should be produced, including translator qualifications, the translation workflow (translation, check, revision), and project management requirements. ISO 21999 focuses specifically on how to evaluate translation quality after production -- it provides error categorization frameworks, severity levels, quality scoring methodologies, and evaluation procedures. In simple terms, ISO 17100 tells you how to produce quality translations, while ISO 21999 tells you how to measure whether you succeeded. Together, they create a complete quality management cycle.
ISO 21999 provides a systematic error categorization framework with five main categories: accuracy errors (mistranslation, omission, addition, untranslated text), linguistic errors (grammar, spelling, punctuation, syntax), terminology errors (incorrect terms, inconsistent terminology, glossary non-compliance), style errors (register, tone, readability, natural expression), and locale/convention errors (date/number formatting, measurement units, cultural adaptation). Each error is also assigned a severity level: critical (could cause harm or serious misunderstanding), major (significantly affects meaning or usability), minor (noticeable but does not significantly affect comprehension), or preferential (alternative rendering, not a true error).
Yes. The error categorization and quality scoring frameworks defined by ISO 21999 can be implemented in most modern CAT tools and TMS platforms. Tools like memoQ, Trados, Phrase (formerly Memsource), XTM, and others support customizable QA frameworks that can be configured to align with ISO 21999 categories and severity levels. This allows reviewers to annotate errors using standardized categories directly within the tool, with automated quality score calculation. TranslationCert can help you configure your existing technology to support your ISO 21999 quality framework.
No, ISO 21999 can be obtained independently. However, the two standards complement each other powerfully. ISO 17100 ensures your production processes are sound (qualified translators, mandatory revision, documented workflows), while ISO 21999 ensures your quality evaluation processes are rigorous and objective (standardized error categorization, quality scoring, evaluation procedures). Many agencies pursue both for the most comprehensive quality credentials. TranslationCert offers bundled packages for agencies seeking multiple certifications.
With TranslationCert, typical certification takes 2-4 weeks. Agencies that already have structured QA frameworks in place (even if not formally aligned with ISO 21999) can often complete the process faster. The timeline includes the initial QA framework assessment (1-2 days), quality framework design and documentation (3-5 days), implementation and evaluator calibration (1-2 weeks), and the certification audit (1-2 days). The key variable is the evaluator calibration phase, which ensures all your reviewers apply the framework consistently.
Evaluator consistency (inter-rater reliability) is a crucial aspect of ISO 21999 implementation. Key practices include: comprehensive training on the error taxonomy and severity definitions; regular calibration sessions where multiple evaluators assess the same translation samples and discuss discrepancies; documented decision guidelines for borderline cases; periodic inter-rater reliability testing to measure and improve consistency; and a feedback loop where calibration data drives refinement of evaluation criteria. TranslationCert provides calibration methodologies and training materials tailored to your specific needs.
Yes. ISO 21999's quality evaluation framework can be applied to any translation output, whether produced by human translators, machine translation, or a combination (MTPE). For MT evaluation, the framework is particularly useful for benchmarking raw MT quality across engines and language pairs, evaluating post-edited MT output against defined quality thresholds, comparing MT quality trends over time as engines improve, and providing objective data for MT engine selection decisions. When combined with ISO 18587 (MTPE), ISO 21999 provides a complete framework for managing and measuring AI-augmented translation quality.