A Test Pass From Build to Sign-off
Localization testing works best as a planned sequence, not a last-week scramble. Below is a typical pass for a console and PC release in several languages, with the failure each step is designed to stop.
The ten steps
Teams rename and merge these steps, and smaller projects compress several into one week. The order, however, holds almost everywhere, because each step depends on the stability created by the one before.
Kickoff and scope
Producer, localization manager and QA lead agree on languages, platforms, test hours per language, bug categories, severity definitions and the exit criteria for sign-off. A word count and a content map (dialogue, UI, tutorials, store text, legal screens) set the size of the job.
Prevents: untested content types, such as store pages or accessibility menus, discovered after launch.
Loc kit, glossary and references
Testers receive the glossary, style guides per language, character bios, a register sheet, a debug or cheat guide, and access to the source-language reference build. Without cheats, reaching every string in a forty-hour game is not realistic.
Prevents: inconsistent terminology and endgame text that nobody ever sees before release.
Pseudo-localization
Before real translations arrive, engineers replace English with stretched, accented placeholder text such as "[Ŝåvé Ĝåmé ~~~]". It exposes hardcoded strings, missing font glyphs and layouts that cannot grow by 30–40%.
Prevents: truncated buttons and untranslatable hardcoded text, at a stage when fixes are cheap.
String freeze
The team agrees that source text stops changing, except through a tracked change process. Translation and testing can then work on a stable target.
Prevents: late English edits that ship untranslated or out of sync with voice-over.
Localized smoke test
The first build with translations is checked quickly in every language: does it boot, can the language be selected, does the font load, do save files work after switching language?
Prevents: wasting tester days on a build that crashes in Korean or shows only boxes in Polish.
Linguistic pass
Native testers play through and read every reachable string in context, logging mistranslations, register problems, plural and gender errors, untranslated text and truncation with screenshots and back-translations.
Prevents: literal idioms, wrong honorifics, "1 items" and visible placeholder code.
Functional localization pass
Testers check what depends on locale rather than language: date, time, number and currency formats, keyboard and IME input, sorting of names, right-to-left layout, subtitle timing and lip-sync, and audio language switching.
Prevents: "12/03" read as the wrong date, garbled Arabic, and Japanese players unable to type names.
Cultural review
Market reviewers go through flagged assets and key scenes: symbols, gestures, maps, names, humour, and content that may affect age ratings.
Prevents: offensive emotes, restricted imagery and rating surprises.
Triage, fixes and regression
Bugs are prioritised by severity, fixed by translators or engineers, and verified in a new build. Testers re-run affected areas, because a fix for German line length can easily break French.
Prevents: fixes that never reach the build and new bugs introduced by old fixes.
Compliance and sign-off
Platform terminology and system messages are checked against the console makers' requirements, legal and age-rating screens are verified, and each language lead signs off against the exit criteria agreed at kickoff.
Prevents: failed certification submissions and launch delays.
How much time to plan
Estimates vary with genre. Text-heavy role-playing games may need several full passes per language, while a puzzle game with a few hundred UI strings can be covered in days. A common planning approach is to estimate hours per language from word count and play-through length, then add a fixed regression buffer of around a quarter to a third of the first-pass time. Underestimating regression is the most frequent planning mistake we see.
Live games never fully sign off
For live-service games, the pass repeats with every patch, season or event. Teams usually keep a small core of language testers who know the game, run a short pass on new content, and spot-check older areas, since engine updates can reintroduce font or layout bugs in text that was correct for years.