You’re right about the difficulties of language translations for surveys and other measurement methods.

There are methods that currently exist (and continue to be developed) for checking such things, however. Measurement Equivalency/Invariance (MEI) procedures using either Structural Equation Measurement Models or Item Response Theory item parameters examine if the same things are being measured between subgroups. In other words, if I’m trying to gauge employee satisfaction for a multinational organization, MEI techniques can tell me if the items are getting at satisfaction levels in all countries involved, rather than unwanted error due to translation issues or cultural differences.