A translation memory (TM) contains a database of translations that translators can share and reuse. Using a TM reduces translation costs and increases consistency between translations. The main function of a TM is to find an exact match for sentences in a source document, or failing that to find a near match, and to then provide suggested translations. A TM has just one source language but may contain translations into many languages.
The business purpose of TMs is to leverage (take advantage of) previously translated text.
You get leverage when a translator finds a match in a TM for a segment of source text . You measure leverage by counting the number and type of matches.
A translator uses a TM to find a match between a segment of presented text and some source language text in the TM. If the presented text exactly matches the source in a translation unit (TU), the TM provides an exact match translation. If the presented text is similar, but not identical, to some source text, the TM suggests one or more translations based on a fuzzy match.
Because a TM holds descriptive data for each TU (for example the date, author and status of the translation), the translator can evaluate an offered translation and can also choose between different offered translations.
New or changed translations can be added to a TM. Studio (with its default settings) updates the TM with a translation when that translation is approved (confirmed).
A Local TM is stored as a file and can only be accessed by one person at a time but a server-based TM is stored in a database and can be accessed by many people at the same time. Also, a local TM has only one target language but a server-based TM can contain multiple target languages.