At CLEAR Global / Translators without Borders we assess quality using the harmonized Dynamic Quality Framework - Multidimensional Quality Metrics (DQF-MQM) error typology.
We use a simplified version of this error typology with five error categories. Each category includes specific issues that should be considered when we work on a project.
Find below more information on how we score for accuracy.
Accuracy. We want to make sure the target text communicates the meaning of the source text correctly and precisely, without adding or removing any information.
Here are some of the most relevant accuracy issues:
- omission of a text from the source or addition, for example:
omission happens when a “blue cat” becomes just a “cat”, and addition when it turns to be a “fluffy blue cat”
- mistranslation: when the target meaning does not represent the source:
a “blue cat” is a out of sudden a “blue computer assisted-translation” when you wanted to talk about an animal
- over-translations, where we are adding a meaning, or under-translations where we are not including enough meaning from the source
we have an over-translation when a “blue cat” is a “blue kitten” and an under-translation, if it turns to be a “blue animal”
- issues when something should have been translated but was left untranslated
“niebieski kot” instead of a “blue cat” could be a name of a brand instead of an animal, and mislead the readers
Note: there are situations when we want to leave something untranslated. This might apply to proper nouns, such as names of brands, companies, organizations, etc. In case of doubt, contact your project officer for clarification.
How to score accuracy?
Make sure you are classifying the issues in the correct category and consider the impact those issues have on the target text. Are they significant? How do they impact the understanding by the target audience?
Keep in mind you are scoring your colleague’s whole chunk or translation. It is important to provide a fair assessment of the complete work.
Make sure you praise exceptional work! 
It is also important to leave a comment about the category together with the score, so your peer can understand the feedback. This is particularly vital when leaving low scores. If you were to receive that score yourself, you would surely appreciate a colleague who also took the time to be helpful!
We know this is not easy, so we put together a few ideas on what each level might mean to guide you on how to decide on a score:
5 Stars: Excellent. The target text communicates the complete information of the source. There might be a minor slip or two, but no major issues with accuracy.
4 Stars: Good. Overall, accuracy was consistent. There are some minor slips, and occasional issues with accuracy choices.
3 Stars: Satisfactory. The translation was accurate for the most part but there were a few issues with meaning.
2 Stars: Needs work. There were numerous mistranslations that impacted meaning in the target text.
1 Star: Poor. The target does not communicate the meaning of the original text correctly and precisely, omits information or adds extra details.
This is meant as guidance, and there will be situations where not all conditions apply. You should always look at the specific case in front of you to decide!