The tracker
Jira is common at larger studios; others use YouTrack, Redmine, Azure DevOps or a vendor's own platform. What matters is one database for all languages, filterable by language, category, severity and build.
Desk 4 of 4 · Tools and norms
Checklists, Trackers and Norms
Talent finds bugs; tools make the findings comparable, traceable and fixable. This desk covers what sits on a reviewer's screen besides the game.
Checklists
A checklist does not replace judgement; it stops tired testers from forgetting the boring parts. Most teams keep one general list and one per language. A compact general list:
- Language selection works from the first boot and persists after restart.
- All fonts render every character used in the language, including uppercase accented letters.
- No string is truncated, overlapping or running outside safe areas on the smallest supported screen.
- Every placeholder resolves; numbers, dates and currencies follow the locale.
- Platform terms, button names and system messages match the console maker's wording.
- Subtitles match spoken audio and stay on screen long enough to read.
- Legal, privacy, age-rating and credits screens are present and translated.
- Store page, achievements or trophies, and patch notes are included in scope.
Trackers and reports
Whatever the software, the shape of a good bug is the same.
The fields
String ID, language, build number, location in game, steps to reach it, screenshot or video, current text, suggested text, back-translation, category, severity. Missing any of the first four usually means the bug bounces back.
The capture
Console capture tools, a phone camera for handheld modes, and a naming rule for files that includes language and ID. Video beats screenshots for subtitle timing and animated UI.
Severity and quality models
Severity tells producers what to fix first. Many game teams use four levels: critical (crash, blocker, legal or offensive content), major (meaning wrong, story-critical text, unreadable UI), minor (grammar or style that does not mislead) and preferential (a valid alternative wording).
For measuring overall quality, the industry borrows from translation research. The LISA QA Model, from the former Localization Industry Standards Association, introduced weighted error categories. SAE J2450, originally written for automotive service documents, offered a simple scored metric. Today the Multidimensional Quality Metrics framework (MQM) provides a detailed, customisable error typology, and many vendors build scorecards on it.
File formats testers meet
Testers rarely edit source files, but they should recognise them when reporting. XLIFF is the XML exchange standard for translatable content; gettext PO files appear in many open-source engines; spreadsheets and JSON remain common in smaller studios. CSV exports deserve caution: a wrong encoding turns accented characters into garbage, and a stray comma can shift every column after it.
The norms on the shelf
| Reference | What it covers | Where testers use it |
|---|---|---|
| ISO 17100 | Requirements for translation services, including translator competence and mandatory revision. | Vendor selection; confirming text was revised before testing. |
| ISO 18587 | Full post-editing of machine translation output. | Projects that use MT for large volumes of item or quest text. |
| Unicode Standard | Character encoding for virtually all writing systems. | Diagnosing missing glyphs and encoding corruption. |
| Unicode CLDR | Locale data: plurals, dates, numbers, currencies, sorting. | Checking plural forms and formats per language. |
| UAX #9 (Bidi Algorithm) | How mixed left-to-right and right-to-left text is displayed. | Arabic and Hebrew UI, especially with numbers and names. |
| ICU MessageFormat | Syntax for plural, gender and select variants in messages. | Understanding why a sentence variant does not appear. |
| MQM | Error typology and scoring for translation quality. | Quality scorecards and vendor reporting. |
| Sony TRC, Xbox Requirements, Nintendo Lotcheck | Platform certification requirements, including terminology. | Compliance check before submission. |
A word on automation
Automated checks are excellent at what is mechanical: length limits, missing placeholders, untranslated strings, inconsistent terms against a termbase, and screenshots captured across languages by scripted runs. They cannot judge whether a joke works or a queen is addressed properly. The most effective teams automate the mechanical layer so that human testers spend their hours on meaning.