Currently, the only metric we use to evaluate the quality of published translations is whether they were deleted or reverted.

However, only very bad translations are likely to be deleted or reverted, so this metric doesn't really help distinguish between mediocre translations (which place an extra burden on other community members) and good ones (which do not). It would be nice to explore alternatives that provide a more nuanced understand of the quality of a translation.

This project is related to our more general desire for more nuanced metrics for article quality, but measuring the quality of translations is likely to have some significant differences. For example, normally we would expect the presence of internal links, citations, images, and templates to be positively associated with quality, but ContentTranslation automatically uses the source article to populate the translation with these rich elements, which would invalidate this association.