Your TMS QA module validates strings against rules. Loqalit validates the rendered product. Here’s which error types each catches, and why you need both layers.
Updated October 2026
Modern TMS platforms ship solid automated QA. Typically they will catch:
This is genuinely valuable, it runs automatically, and it costs you nothing extra. If you are not using it, turn it on before you evaluate anything else.
Every check in that list operates on the segment, comparing it against the source or a rule. That design has a hard boundary: it cannot evaluate anything that only exists after rendering.
Which means it does not catch:
That last one is worth pausing on. Anything that never entered your TMS cannot be checked by your TMS. Hardcoded strings, third-party widget text, error states, image text, these are among the most commonly shipped localization defects, and they are invisible to string-level QA by definition.
New to rendered-product QA? Read the guide: Localization QA for live websites and web apps.
| Error type | TMS QA | Loqalit |
|---|---|---|
| Missing translations | Yes | Yes |
| Terminology vs glossary | Yes | Yes |
| Placeholder mismatches | Yes | Yes |
| Text overflow in the rendered UI | No | Yes |
| Truncation and clipping | No | Yes |
| Layout breaks and overlap | No | Yes |
| RTL mirroring errors | No | Yes |
| Context-wrong but segment-correct translations | No | Yes |
| Cross-screen terminology conflicts | No | Yes |
| Hardcoded / untranslated strings in the build | No | Yes |
| MQM-scored severity across the product | No | Yes |
Loqalit integrates with Crowdin and Phrase, so findings can flow back into the workflow where the fix happens. Your TMS remains the system of record for translation. Loqalit becomes the verification layer between translated and shipped.
The mental model that works: your TMS checks the text you sent it. Loqalit checks the product your users get.
TMS QA checks are valuable and you should have them switched on. They catch missing translations, terminology violations against a glossary, placeholder mismatches, and formatting discrepancies. But every one of those checks operates on the segment, comparing it against the source or a rule, so none of them can evaluate anything that only exists after the page is rendered.
Text that overflows the actual button despite fitting a declared character limit, truncation and clipping at specific viewports, layout breaks and overlap, RTL mirroring errors, translations that are correct for the segment but wrong for the UI element they label, and terminology that matches the glossary but contradicts an adjacent screen. It also cannot check any string that never entered the TMS at all, hardcoded text, third-party widget copy, and image text.
No. Your TMS remains the system of record for translation. Loqalit is the verification layer between translated and shipped. The two check different things: your TMS checks the text you sent it, Loqalit checks the product your users get.