TRANSLATION QUALITY & GOVERNANCE

Translation Quality · Program Governance

Translation Quality Management Guide

Learn how to define, measure, review, and improve translation quality across languages, teams, content types, and recurring projects.

Stepes Resource Guide Quality requirements, review models, measurement, and corrective action
Read the key takeaways ↓

Quality requirements

Content purpose · Audience · Risk

Defined control and evaluation

  • Translation review
  • Issue resolution
  • Approved release
  • Quality requirements
  • Risk-based review
  • Evaluation criteria
  • Continuous improvement

Key Takeaways

This guide connects requirements, qualified resources, terminology, workflow controls, evaluation methods, accountability, and continuous improvement into one practical quality-management model.

Define acceptance criteria based on the content’s purpose and consequences of errors, not a vague request for “high quality.”

Evaluate linguistic issues using consistent categories, severity levels, representative samples, and reviewer instructions.

Close the loop by correcting content, terminology, workflows, and root causes—not only individual translated sentences.

Translation Quality Is a Managed Process

Quality is not determined by a final proofreading pass alone. A reliable program starts by defining what the translation is for, who will use it, and which errors would matter. It then selects appropriate language resources, production methods, review controls, and evidence of approval.

Fitness for Purpose

Choose criteria for accuracy, completeness, terminology, style, usability, and relevant technical requirements.

Ownership and Accountability

Set clear responsibilities for content owners, translators, reviewers, in-country stakeholders, and final approvers.

Repeatability

Use shared language assets, documented decisions, and recurring measures rather than relying entirely on individual reviewer preferences.

Six Foundations of an Enterprise Translation Quality Program

The following elements are useful whether translation is fully human, AI-assisted, or produced through a controlled hybrid workflow.

Practical considerations

  • Content Classification Segment material by purpose, audience, subject matter, public visibility, and consequences of errors.
  • Quality Specification Define measurable acceptance criteria, review depth, terminology requirements, and the scope of format or functional testing.
  • Qualified Resources Match linguistic, technical, and domain expertise to the content and language pair.
  • Language Assets Use approved glossaries, translation memory, writing guidance, and source-content references.
  • Review and Escalation Specify who reviews issues, how disagreements are resolved, and which changes require specialist approval.
  • Measurement and Improvement Track error trends, corrections, review effort, recurrence, and actions taken to prevent similar issues.

RISK-BASED QUALITY

Choose Review Controls by Content Risk

An internal knowledge article and a clinical or safety-critical instruction should not automatically follow the same approval path. A risk-based model matches controls to consequences while maintaining baseline accuracy for every content type.

Lower-Risk Internal Content

AI-assisted translation with terminology checks, automated QA, and selective human sampling may fit some internal uses.

Customer-Facing and Technical Content

Use professional linguistic review, context, product terminology, and relevant technical checks.

Legal, Financial, Medical, or Regulated Content

Consider qualified specialist review, independent revision where required, defined approval authority, and documented evidence.

High-Visibility Creative Content

Evaluate tone, intent, local audience response, and brand suitability with human-led review or transcreation.

Calibrate Reviewers and Resolve Disagreements

Two qualified reviewers may interpret a tone issue differently while agreeing that a missing warning is serious. A quality program needs definitions, examples, and a dispute-resolution path so assessments reflect agreed requirements rather than individual preferences.

Reviewer Instructions

Provide language, domain, audience, terminology, error categories, and scoring guidance before evaluation.

Severity Calibration

Discuss examples of minor, major, and critical errors and how they affect acceptance.

Dispute Resolution

Define the authority for terminology decisions, ambiguous source passages, and contested quality findings.

Evidence and Records

Retain reviewed source/target versions, issue descriptions, corrections, and approval decisions appropriate to the engagement.

Example: Define a Translation Quality Specification

For a global support knowledge base, the quality brief might identify a target audience, preferred technical terminology, required translation coverage, escalation rules for uncertain instructions, an approved review sample, and formatting acceptance checks. That is more useful than simply requesting “high quality.”

Worked example

Use these decision stages to structure a review of the situation described above.

  1. 01
    Define

    Purpose, risks, and acceptance criteria

  2. 02
    Evaluate

    Linguistic issues and severity

  3. 03
    Improve

    Corrections, feedback, and updates

Build and Operate a Quality Management Cycle

A quality process should create actionable feedback, not just a pass-or-fail label.

  1. 01
    Specify Requirements

    Document intended use, market, languages, terminology, critical content, and approval criteria.

  2. 02
    Prepare and Assign

    Prepare language assets, select qualified resources, define review independence, and communicate requirements.

  3. 03
    Produce With Controls

    Apply the selected translation workflow, consistency checks, and protected-data handling.

  4. 04
    Evaluate and Correct

    Assess meaning, completeness, terminology, style, format, and errors by type and severity.

  5. 05
    Approve and Release

    Resolve material findings and record final approval in the defined client process.

  6. 06
    Learn From Results

    Refresh glossaries, guidance, training, workflow rules, and evaluation samples based on recurring issues.

Use Quality Findings to Improve the Source Process

Translation errors sometimes originate in unclear source writing, unstable terminology, or incomplete product context. Treating every issue as a translator failure misses opportunities to improve the overall system. Quality records should identify whether the root cause lies in source content, language assets, production, review, or formatting.

Source Quality

Improve ambiguous instructions, inconsistent terminology, and unclear references before recurring translation work.

Language Asset Maintenance

Update approved glossaries and translation memory after authorized review decisions.

Process Improvement

Use recurring defects and review delays to refine briefing, resource assignment, QA checks, and escalation paths.

Measure Translation Quality Without Relying on One Score

Error counts can be useful, but they require context. Programs should define the sample, error typology, severity, reviewer instructions, and acceptance threshold before comparing vendors, language pairs, or models.

Error Categories and Severity

Use a consistent issue model such as MQM with relevant error types, severity rules, and clear descriptions.

Representative Sampling

Evaluate content representative of actual projects, languages, difficulty, and layout conditions; avoid cherry-picked sentences.

Agreement and Calibration

Train reviewers on the specification, discuss borderline cases, and assess differences in judgment.

Corrective Actions

Track repeated issues by cause and ownership so terminology, processes, and source content improve.

Match Quality Controls to Content Risk

Use this framework to compare requirements and identify decisions that need an owner before work begins.

Content context Recommended controls to consider Evidence to retain
Internal reference material Terminology guidance, fit-for-purpose evaluation, and sampling where appropriate Scope, sampling method, decision and feedback
Customer-facing technical content Professional linguistic revision, context checks, and protected-value validation Reviewed version, issue log, acceptance signoff
Safety, legal, medical, or regulated material Subject-matter escalation and review steps required by client obligations Approved source, specialist findings and release decision

Common Mistakes to Avoid

  • Comparing raw quality scores without checking sample size, error categories, and severity weighting.

  • Treating preference differences as equivalent to material accuracy or safety errors.

  • Skipping source-content and glossary improvements after recurring issues emerge.

Frequently Asked Questions

What does translation quality management include?

It includes requirements, qualified resources, terminology, translation process controls, review, issue resolution, approval, metrics, and improvement across projects.

Is linguistic quality assurance the same as proofreading?

No. LQA can evaluate a translation against structured error categories, severity, completeness, terminology, and context. Proofreading is a narrower editing step.

Can an MQM score be compared across programs?

Only if the sampling, error categories, severity weights, counting method, and thresholds are compatible. A score without its methodology can be misleading.

Does every translation need the same human review?

No. The appropriate review depth depends on the audience, subject matter, impact of errors, client requirements, and applicable standards or regulations.

What is the difference between automated QA and linguistic review?

Automated QA detects patterns such as numbers, tags, terminology, and missing segments. Qualified human reviewers evaluate meaning, fluency, appropriateness, and nuanced errors.

How do teams improve quality over time?

Record patterns in defects and feedback, update terminology and instructions, calibrate reviewers, verify corrective actions, and monitor whether errors recur.

Review Translation Quality Requirements and Workflows

Share the content types, languages, review challenges, and quality requirements you manage. Stepes can help design suitable translation and validation workflows.