So just how big is the business issue for companies involved in localization? According to research conducted by LISA in 2004 (Lommel 2004), companies involved in localization use TM tools for an average of 7.2 target languages, and the databases of a typical TM user contains approximately 350,000 source segments (typically sentences). This would mean that, subtracting the source language, a typical TM repository might represent a total of just under 2.2 million target-language segments. The number of words per segment was not addressed in the LISA study, but a rough figure of ten words per segment will be assumed for purposes of estimation, yielding a figure of 22 million words in a TM repository. Assuming for purposes of comparison a per-word translation cost to the company of USD 0.20/word, a typical company involved in localization might have more than USD 4 million of information tied up in a given translation tool. Given that not all text is translated into all languages, a more realistic figure might be around USD 2 million, still a substantial investment in the production and maintenance of information.

At an extreme end, some companies have as much as 100 million source segments in TM databases, and localize into thirty or more languages. Using the same method of calculation as before, such a large-scale user of TM tools could conceivably have TM assets worth more than USD 3 billion (3,000,000,000). Not all TM data is created equally, however, and some segments should be considered "garbage - outdated or inaccurate segments that nonetheless remain in the database. Information can be rendered obsolete for any number of reasons, and one major task for managers of translation memory databases is pruning outdated data from the database to increase accuracy and improve performance. In the multilingual realm, a frequent source of data corruption is the failure to update databases to reflect final revisions of translated text. In the absence of appropriately changed segments, the database will contain outdated versions of translated text that may actually have an associated cost for companies since they will need to be corrected repeatedly, in each unique target-language instantiation. When the value of these garbage segments is subtracted, the overall value of the information will decrease somewhat.

The actual numbers themselves are really beside the point, since the real value of multilingual corporate information cannot be objectively and quantitatively calculated on a universally accepted basis. In addition, these figures consider only the cost to produce and maintain the information, not its potential positive business value in sales, service and support. The point of these figures is to show that the value of the information contained in a TM database exceeds the cost of the tools themselves by several orders of magnitude, a fact that makes independence from a specific tool highly desirable.