Microcopy and Tooltip Localization

Quick answer — Microcopy fails on missing context rather than difficulty. A word like Open or Post can be a verb or a noun, and a translator with no screenshot and no character limit guesses wrong about half the time.

Vitra.ai Universe ships product and customer content in every language.

Short strings are the expensive ones

A thousand words of help content is straightforward work. Forty button labels is not, and it will produce more corrections. The reason is that a short string carries almost no information about itself. "Open" could be a command on a button, a status on a badge, or an adjective in a filter. Every language resolves those three differently, and nothing in the string says which it is.

Send the context, not just the string

A string sent without these is not a task, it is a question — and the answer you get back is a guess that reads fine and is wrong often enough to matter.

Length is a design constraint

English microcopy is unusually compact. German runs longer, and a button sized to fit "Save" has nowhere to put "Speichern unter" when a label grows.

Two honest options. Give the component room in the source design, which is the same argument as leaving slack in a banner layout. Or set a character limit and let the translator write to it rather than translate to it — a shorter, different word that fits beats an accurate one that truncates.

Do not reuse strings to save money

Sharing one "Open" across six places looks efficient in the codebase and guarantees at least one of the six is wrong, because a single translation now has to serve six grammatical contexts.

Separate keys per context cost nothing and remove a whole class of defect, and translation memory still reuses the ones that genuinely are identical.

Check it rendered, not extracted

Microcopy problems are invisible in a string file and obvious on a screen.

Run pseudo-localization before translation to find the fixed-width containers, then review the built interface rather than a spreadsheet. Quality control covers wording; truncation needs the rendered view. Screens built into hardware cannot reflow at all — infotainment string translation.

FAQ

Why is microcopy harder to translate than long content? Because a short string carries no information about itself. A word like Open can be a verb, a noun or a status, and nothing in the string tells the translator which, so they guess.

What context should be sent with UI strings? A screenshot or component name, the character limit, the part of speech where ambiguous, whether the string is reused, and the meaning of any placeholder. Without those it is a question, not a task.

Should the same string be reused across the interface? No. Sharing one label across several contexts guarantees at least one is wrong, because a single translation has to serve several grammatical situations. Separate keys per context cost nothing.

How should translated microcopy be checked? On the rendered interface rather than in a string file. Truncation and overflow are invisible in a spreadsheet and obvious on screen, and pseudo-localization finds the fixed-width containers first.