The QA Check Tool runs a set of rule-based checks against a work unit's segments — tags, whitespace, glossary terms, length, and more — and flags anything that looks wrong before you deliver a file. It never blocks delivery on its own; it just points you at segments worth a second look.
💡 Who is this for? This guide is for Vendors and Project Managers who need to run and interpret QA checks on translated content inside the wxrks Editor.
Opening the QA Issues Panel
The QA Check Tool lives inside the wxrks Editor, scoped to whichever work unit (or joined work units) you currently have open.
Open the wxrks Editor for the work unit(s) you want to check.
In the right-hand icon bar, click the QA Issues icon (the warning triangle).
⚠️ If the work units are joined, the panel replaces the single Run QA button with Run for Current (checks only the file you're viewing) and Run for All (checks every joined file).
Choosing Which Checks Run
Click Filters to open QA Settings — a checklist of every available check, grouped by category (General, Tags, Characters, Length, CSV). Each category has its own Select All / Deselect All shortcut, and some checks reveal extra sub-options (like a character limit or an exceptions list) once you turn them on.
You can also Save Filter to name and reuse a specific combination of checks later, or load one back with Saved Filters. Once you've made your selections, click Apply — this closes the panel and immediately re-runs QA with the new settings.
⚠️ Some accounts also have an Enable Smells toggle at the top of this panel, under Execution. That turns on wxrks' separate AI-based quality review, which layers Confidence Scores and "Smells" on top of these checks — see QA Confidence Score and Smells for that feature specifically. Everything else on this page is about the standard, rule-based checks below.
Running a Check and Reading the Results
Click Run QA (or Run for Current / Run for All for joined work units) to start the check. QA runs in the background — you can keep working while it finishes — and the panel shows a RUNNING… state until it's done.
When the run finishes, results appear as a list, one entry per flagged segment, with the segment number, a warning count, and each specific issue found:
Above the results, "The report was run on…" shows when this data was last generated — if you've since edited segments, click Run QA again to refresh it, since results don't update automatically until you re-run.
Filtering what you see
A row of issue type chips appears above the list (for example, TARGET IS SAME AS SOURCE, GLOSSARY, TARGET LENGTH). Click a chip to show only that issue type.
Use Segments with Issue / Segments without Issue to narrow the segment list itself to only flagged or only clean segments.
Click Clear (next to the toggles) to reset every active filter at once.
Jumping to a flagged segment
Click anywhere on a result's header row (or its link icon) to jump straight to that segment in the editor, so you don't have to scroll and hunt for it manually.
Ignoring an issue
If a warning is a false positive, click the eye icon next to it to mark it Ignored — it disappears from the default view. Turn on Show ignored issues at the top of the panel to bring hidden warnings back into view (and click the eye icon again to restore them). You can also ignore every instance of one issue type at once from its chip, using the eye-slash icon on the chip itself.
⚠️ Ignoring an issue is tied to this specific work unit, not to you personally — anyone who opens the same work unit sees the same ignored/visible state. It also isn't remembered automatically the next time the file goes through a different workflow step: if an issue should stay ignored throughout the whole job, the person handling each later workflow needs to ignore it again in their own pass.
Downloading a report
Click Download Report to export the current QA results as a spreadsheet. The export modal lets you choose whether to include ignored issues in the report before generating the file.
Analysis Points Reference
These are the checks available in QA Settings, grouped the same way they appear in the panel. None of them block delivery — they're guidance, not a gate — but each one flags a specific pattern worth a human look. Checks marked on by default already run unless you turn them off in Filters.
General
Check | What it flags |
Warn if whitespace mismatch (on by default) | Leading, trailing, or repeated internal whitespace that differs between source and target. Each sub-case (leading, trailing, multiple spaces) can be toggled separately. |
Warn if a target segment is empty when its source is not empty (on by default) | The translation is blank but the source has content. |
Warn if a target segment is not empty when its source is empty (on by default) | The source is blank but the translation has content. |
Warn if a target segment is the same as its source (on by default) | The translation is identical to the source. Only fires when the source has at least one letter or digit, so numeric-only or symbol-only segments don't get flagged. |
Warn on doubled words | The same word repeated back-to-back (for example, "the the"). Off by default; has an exceptions list for languages where repeating a word is normal grammar. |
Warn if repetitions are different (on by default) | The same sentence repeats in the source but was translated two different ways. |
Warn if glossary mismatch (on by default) | A term from a linked Glossary appears in the source but its approved translation isn't found in the target. |
Warn if there are potential spelling errors | Off by default; separate from the Editor's own live spell-check underlines. |
Warn if non-breaking spaces mismatch / Warn if narrow non-breaking spaces mismatch (on by default) | A non-breaking or narrow non-breaking space in the source has no matching one in the target, or vice versa. |
Warn if invalid characters in the XML (on by default) | The target contains a character that would corrupt the file's XML structure on export. |
Tags
Check | What it flags |
Warn if there are tag differences between source and target segments (on by default) | A tag present in the source is missing from the target (Missing tag), or a tag appears in the target that isn't in the source (Extra tag). |
Try to guess opening/closing types for placeholder tags (on by default) | Lets the checker infer whether a generic placeholder tag functions as an opening or closing tag, so it can also be checked for correct pairing. |
Enforce strict order of the tags (on by default) | Tags appear in an unexpected sequence in the target compared to the source (Suspect tag). |
Warn if custom style tags have mismatches | Off by default. Validates target-only formatting tags (bold, italic, underline, strikethrough) added directly in the Editor. |
Characters
Check | What it flags |
Warn if a character is not included in the ISO-8859-1 character set encoding | Off by default. Useful when the target file format requires a legacy single-byte encoding. |
Warn if some possibly corrupted characters are found in the target entry (on by default) | A pattern typical of encoding corruption (for example, a UTF-8 file that was opened as ISO-8859-1). Catches common cases, not every possible one. |
Length
Check | What it flags |
Warn if a target is longer than N chars | Off by default; the character limit (255 by default) is editable in the same row. Tags don't count toward the total. |
Warn if a target is disproportionately longer than its source / …shorter than its source (both on by default) | The target's length, as a percentage of the source's, falls outside the configured thresholds — useful for catching truncated or duplicated translations without a hard character cap. |
Warn if a source or target text unit does not fit its ITS storage size property | Off by default. For XLIFF files that declare an ITS storage size limit — checks the actual byte length against it. |
CSV
Check | What it flags |
Warn if a target has extra double-quotes (on by default) | Extra |
Warn if a target has extra semicolons (on by default) | Extra |
Warn if a target has line breaks mismatches (on by default) | Missing or extra line breaks compared to the source. |
Important Behavior and Edge Cases
QA never blocks delivery. Every check here is advisory — you can complete or deliver a task with warnings still showing.
Results don't refresh on their own. Editing a segment after a run doesn't update its warnings until you click Run QA again.
Ignoring is per work unit, not per user or per workflow. See the note under "Ignoring an issue" above — anyone opening the same work unit sees the same ignored state, but it isn't inherited automatically by the file's next workflow step.
Corrupted-character detection is best-effort. It catches common encoding-corruption patterns, not every possible one.
Related Articles
QA Confidence Score and Smells — the AI-based quality review that layers on top of these checks.
Quality Module — how QA results roll up across a whole project.
Quality Dashboard — tracking quality trends over time.
How to configure custom review categories — setting up manual review categories alongside automated QA.