When Translation Memory software is used for translation, the first step is analysis of the source text. For each source word/phrase that the software does not find a target language word/phrase that is close to an accurate translation (above a set threshold of equivalence, generally 60 or 50%), the software returns a “no match” result.