LISA QA Model and translation quality evaluation

The LISA QA Model is a quality evaluation model historically used in translation and localization projects. Unlike ISO 17100, which focuses on the translation service process, the LISA model is oriented towards the translated product: the final text, its suitability, its errors and its level of quality according to defined criteria.

This difference matters. A translation conforming to ISO 17100 requires a qualified translator, independent revision, project management and final verification. The LISA QA Model, by contrast, helps classify errors and assess the quality of a translation already produced.

For that reason, both approaches can complement each other, but they should not be confused.

What the LISA QA Model is

The LISA QA Model is a quality control methodology that allows a translation to be reviewed through error categories and severity levels.

Its use has been especially linked to the localization industry, software, technical documentation and multilingual projects in which the client needs to measure text quality in a more structured way than through a simple reading.

The model is based on a practical idea: not all errors have the same impact.

A minor typo in a secondary sentence does not have the same severity as a terminology error in a technical instruction, an omission in a contractual clause or a wrong translation in a safety message.

Process quality and product quality

Process quality and product quality are two different levels.

ISO 17100 mainly deals with the process. It defines how the service should be organized: prior analysis, qualification of professional translators, independent revision, translation project management, technical resources, handling of comments and project closing documentation.

The LISA QA Model deals with the product. It helps answer another question: what errors does the delivered translation contain and what impact do they have?

An agency can apply a sound process and, in addition, use product evaluation models when the client needs a specific measurement of quality.

How error-based evaluation works

In an evaluation based on the LISA QA Model, the reviser identifies errors and classifies them.

Usually, two aspects are assessed: type of error and severity of the error.

The type of error indicates what has failed. Severity assesses the impact.

For example, an error may be terminological, grammatical, formatting-related, coherence-related, an omission or a problem of contextual suitability. It may also be considered minor, major or critical depending on whether it affects the final use of the text.

This approach makes it possible to move from a general opinion about a translation to a more documented assessment.

Common error categories

The exact categories can be adapted to the project, but translation evaluation models usually analyse aspects such as accuracy of meaning, terminology, omissions, unjustified additions, grammar, spelling, punctuation, style, register, coherence, formatting, tags or variables, adaptation to the destination country, compliance with client instructions, readability and functionality in context.

In website translation, for example, buttons, forms, metadata, truncated texts, links, calls to action and interface strings can also be evaluated.

In technical documentation or software, tags, commands, error messages, variables and elements that must not be translated may also be reviewed.

Severity of errors

Severity classification is one of the most useful elements.

A minor error may affect style, a minor preference or a detail that does not prevent the text from being used.

A major error may alter the meaning, affect terminology or hinder understanding.

A critical error may make the text dangerous, legally problematic, technically incorrect or unusable for its intended purpose.

This grading makes it possible to prioritize corrections. The issue is not only to count errors, but to understand their real effect.

LISA QA Model in localization

The LISA model has been widely used in localization projects.

Localization requires the reviewer to assess not only language, but also the adaptation of the content to the market and to the technical environment in which it will be published.

In software, applications, websites or digital documentation, the evaluation may include out-of-context text, broken variables, damaged tags, character limits, unclear messages, menu errors, inconsistencies between screens, formatting problems, units or dates that have not been adapted and translations that do not fit into the interface.

That is why the LISA QA Model is particularly suitable for projects in which the translation has to function within a product or digital environment.

Difference between ISO 17100 revision and LISA evaluation

ISO 17100 revision is part of the normal translation service process. It is carried out by a person other than the translator and consists of comparing source and target content in order to check the suitability of the text for its intended purpose.

Evaluation with the LISA QA Model is more analytical. It can be used to audit a translation, compare providers, review a sample, measure errors or establish an acceptance threshold.

It does not necessarily replace ISO 17100 revision. It can be added when the client needs a more quantifiable evaluation or when the project requires specific acceptance criteria.

When the LISA QA Model can be useful.

This type of evaluation can be useful in software localization, application translation, technical documentation, multilingual manuals, website projects, recurring translations, comparison of suppliers, review of third-party translations, linguistic audits, projects with quality thresholds, control of large volumes and sample validation.

It can also help detect patterns. If many errors belong to the same category, the problem may not be in one specific sentence only, but in a translation memory, an incorrectly defined instruction or unapproved terminology.

Relationship with SAE J2450 and ISO 5060

The SAE J2450 model is also based on error classification, but it has a more specific field of use: evaluation of translations in the automotive sector.

The LISA QA Model has been used more in localization and multilingual digital content projects. SAE J2450 is more commonly associated with automotive documentation and technical texts in that sector.

Both models help evaluate the translated product, while ISO 17100 regulates the service provision process.

Relationship with ISO 5060.

Evaluation of the quality of the translated product has developed significantly in recent years, especially because human translation, machine translation, post-editing and quality control systems now coexist.

In that context, ISO 5060 is relevant because it focuses on evaluation of translation output, including human translations, post-edited machine translations and unedited machine translations.

In a clear architecture of standards and models, ISO 17100 regulates the translation service process; ISO 18587 relates to post-editing of machine translation; ISO 5060 is oriented towards evaluation of the translated product; and LISA QA Model and SAE J2450 function as error evaluation models or metrics in specific contexts.

Evaluation and machine translation

Machine translation has made objective product evaluation more important.

A machine-generated text can sound fluent and still contain errors of meaning, terminology or suitability. That is why a superficial reading is not enough.

When the service consists of machine translation post-editing, the specific reference is ISO 18587, not ISO 17100.

However, if the objective is to evaluate the final product, error classification models or evaluation-oriented standards can be used, depending on the type of project.

Limitations of the LISA QA Model

The LISA QA Model should not be applied mechanically.

A poorly adapted error table can produce results of little practical use. Not all projects need the same categories. Not all texts should be evaluated with the same level of severity either.

An advertising campaign, a technical manual, a contract and a software interface involve different risks.

The evaluation must be adapted to the purpose of the text, the audience, the sector and the agreement with the client.

Quality evaluation and translation rates

Formal quality evaluation takes time and requires qualified professionals.

For that reason, it should be distinguished from a standard revision included in the translation process. If the client requests a linguistic audit, an error report, a scored evaluation or a comparison between suppliers, that service should be quoted separately.

When comparing translation rates, it is advisable to know whether the quotation includes only translation, translation with revision, quality evaluation, desktop publishing, localization or an error report.

Not all controls have the same scope.

LinguaVox and quality evaluation

LinguaVox can apply professional translation processes under ISO 17100 and, when the project requires it, review third-party translations, prepare quality reports, control terminology, work with translation memories and evaluate multilingual texts according to criteria agreed with the client.

This approach is useful in technical documentation, website localization, software, medical texts, legal documentation, automotive translation and recurring multilingual projects.

Request a translation review or evaluation quote

Frequently asked questions about the LISA QA Model

What is the LISA QA Model?

It is a quality evaluation model that classifies translation errors by type and severity. It has been used especially in localization, software, technical documentation and multilingual projects.

Is the LISA QA Model the same as ISO 17100?

No. ISO 17100 regulates the translation service process. The LISA QA Model is used to evaluate the translated product through error categories and severity levels.

What is the purpose of classifying errors?

It makes it possible to know what type of problems a translation contains and what impact they have. A minor typo is not the same as a serious omission or a critical terminology error.

Can the LISA QA Model be used to review third-party translations?

Yes. It can be used to audit translations already produced, compare suppliers, review samples or prepare quality reports with more structured criteria.

What is the difference between the LISA QA Model and SAE J2450?

Both classify errors, but SAE J2450 is more closely linked to technical translation in the automotive sector. The LISA QA Model has been used more in localization and multilingual digital projects.

Does ISO 5060 replace the LISA QA Model?

Not necessarily. ISO 5060 provides a more current framework for evaluating the quality of translated output. The LISA QA Model remains useful as a historical or practical reference in certain localization and error-control environments.