The Translation quality dashboard in Lokalise Analytics provides insights into the quality and review of translated content across your projects.
This article covers the reports, metrics, and filters available in the Translation quality dashboard, including AI Scoring analytics. For an overview of Lokalise Analytics and the other available dashboards, see Lokalise Analytics.
Translation quality is only available for completed languages within tasks.
Introduction
Use the Translation quality dashboard to analyze review effort and editing activity across languages, workflows, translation methods, translators, and reviewers.
The available reports help you monitor:
post-edit rate
average edit distance
review effort distribution
language quality patterns
workflow performance
translator and reviewer trends
cases where reviewers had to create new translations or retranslate existing content
These insights can help you identify workflows that require additional review, compare translation quality across methods and languages, detect inconsistent review patterns, spot tasks with New text or Retranslation activity, and monitor quality trends over time.
To access the Translation quality dashboard, open the corresponding tab on the Analytics page.
Filtering data
You can filter translation quality analytics data using the following filters:
Date — select the time range for the displayed analytics data.
Time grouping — group data by week, month, or quarter.
Project ID — filter results by specific Lokalise project IDs.
Project — filter data by project name.
Target language — filter analytics data by target language.
Task ID — display analytics data for specific tasks.
Task — filter results by task name.
Translation method — filter data by translation method, such as human translation, AI/MT, or translation memory.
Translation task workflow — filter results by workflow type.
Translator — filter data by contributor responsible for translations.
Reviewer — filter data by contributor responsible for review activity.
These filters can be combined to analyze translation quality trends across projects, workflows, languages, and contributors.
Quality overview
The Quality overview section provides a high-level summary of review activity and translation quality for the selected period.
Review and edit general info
These metrics help you understand how much content was reviewed, how much reviewer effort was required, and whether translation quality is improving or declining over time.
Metrics include:
Translations reviewed — the total number of translations reviewed during the selected period. This includes review tasks where reviewers created new translations due to missing source content or retranslated existing content.
Words reviewed — the total number of source words reviewed during the selected period. This includes review tasks where reviewers created new translations due to missing source content or retranslated existing content. Use this together with translations reviewed to understand overall review workload.
Post-edit rate — the percentage of translations that required edits during review. Lower values usually indicate higher translation quality. Values below 30% are shown as healthier in the dashboard. Period-over-period change shows whether quality is improving or declining over time.
Avg. edit distance — the average edit extent as a percentage of segment length. Lower values indicate fewer reviewer changes. Light edits are below 15%, medium edits are between 15–30%, and heavy edits are above 30%. Use this together with post-edit rate to understand edit intensity.
Each metric also shows how the current period compares to the previous comparable period, helping you monitor quality and review workload trends over time.
Translation effort summary
The Translation effort summary chart shows reviewed translations grouped by edit effort.
Use this chart to distinguish between standard review effort and cases where review turned into translation work. Accepted means no edits were needed, while Light edit, Medium edit, and Heavy edit indicate increasing levels of reviewer changes. New text and Retranslation mean reviewers created new translations instead of editing existing ones.
Categories include:
Retranslation — translations that were retranslated during review.
This means the reviewer created a new translation because the source content changed after translation.
Accepted — translations accepted without edits.
Light edit — translations with minor edits, below 15% edit distance.
Medium edit — translations with moderate edits, between 15–30% edit distance.
Heavy edit — translations with significant edits, above 30% edit distance.
New text — translations treated as newly created text during review.
Higher shares of Heavy edit, New text, or Retranslation may indicate quality, context, or workflow issues that require further investigation.
Translation effort by workflow
The Translation effort by workflow chart shows reviewed translations grouped by workflow type.
Use this chart to understand which translation and review workflows generate the most review activity and where quality or process improvements may be needed. This includes review tasks where reviewers created new translations due to missing source content or retranslated existing content.
Workflow examples include:
Pro AI translation → Human translation — translations created with Pro AI and then handled by human translators.
Human translation (no previous translation in Lokalise) — human translations created without a previous tracked translation in Lokalise.
Human translation → Human translation — translations passed through multiple human translation steps.
Human translation → Review — human translations that were later reviewed.
Review (no previous translation in Lokalise) — review activity without a previous tracked translation in Lokalise.
Review → Review — translations passed through multiple review steps.
Higher shares for a specific workflow may indicate where most review effort is concentrated.
Quality trends
The Quality trends chart shows how post-edit rate and average edit distance change over time.
Use this chart to monitor whether translation quality and review effort are improving, declining, or staying stable across the selected period.
Metrics include:
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Avg. edit distance — the average edit extent as a percentage of segment length. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Rising trends may indicate declining translation quality or increasing review effort. Stable or decreasing trends usually suggest healthier translation workflows.
Quality by translation method
The Quality by translation method chart compares translation quality metrics across different translation methods.
Use this chart to identify which methods require more review effort and where additional context, workflow improvements, or quality checks may be needed.
Metrics include:
Translations reviewed — the number of reviewed translations for each translation method.
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Avg. edit distance — the average edit extent as a percentage of segment length. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Higher post-edit rate or average edit distance values may indicate lower initial translation quality or workflows that require more reviewer effort.
Edit effort by translation method
The Edit effort by translation method chart shows how review and translation effort is distributed across translation methods.
Use this chart to compare how often translations from each method are accepted as-is, require light, medium, or heavy edits, or require reviewers to create new translations. Larger shares of Heavy edit above 30%, New text, or Retranslation may indicate lower translation quality, insufficient context, or workflows that require new translations.
Categories include:
Accepted — translations accepted without edits.
Light edit — translations with minor edits, below 15% edit distance.
Medium edit — translations with moderate edits, between 15–30% edit distance.
Heavy edit — translations with significant edits, above 30% edit distance.
New text — translations treated as newly created text during review.
Retranslation — translations that were retranslated during review.
This means the reviewer created a new translation because the source content changed after translation.
Larger shares of Heavy edit, New text, or Retranslation may indicate lower initial translation quality, insufficient context, or workflows that require optimization.
Review volume by effort
The Review volume by effort chart shows review volume over time, grouped by edit effort.
Use this chart to understand whether review workload is shifting toward lighter or heavier edits. Healthy growth is typically reflected by Accepted and Light edit volumes growing proportionally with overall review activity.
Rising shares of Heavy edit, New text, or Retranslation may indicate declining translation quality or review tasks that require reviewers to create new translations.
Categories include:
Accepted — translations accepted without edits.
Light edit — translations with minor edits, below 15% edit distance.
Medium edit — translations with moderate edits, between 15–30% edit distance.
Heavy edit — translations with significant edits, above 30% edit distance.
New text — translations treated as newly created text during review.
Retranslation — translations that were retranslated during review.
Healthy growth is usually reflected by Accepted and Light edit volumes growing proportionally with overall review activity. Rising shares of Heavy edit, New text, or Retranslation may indicate declining translation quality, insufficient context, or scaling issues.
Language evaluation
The Language evaluation tab helps you compare translation quality and review effort across language pairs, translators, reviewers, and translator-reviewer pairings.
Use this section to identify language pairs that require more corrections, spot inconsistent review patterns, compare contributor performance, and understand where additional context or workflow improvements may be needed.
Language pair difficulty ranking
The Language pair difficulty ranking chart compares language pairs by review effort and translation quality indicators.
Use this chart to identify language pairs that may require more reviewer corrections, additional context, or longer turnaround times.
Metrics include:
Avg. edit distance — the average edit extent as a percentage of segment length. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Translations reviewed — the number of reviewed translations for the language pair.
Higher values for both post-edit rate and avg. edit distance may indicate language pairs that require more review effort or quality improvements. The number of translations helps you understand whether the results are based on a meaningful review volume.
Translation edit rate by language
The Translation edit rate by language chart shows how review effort is distributed across edit types for each language pair.
Use this chart to identify language pairs that are frequently accepted without changes, require reviewer edits, or include cases where reviewers created new translations. Larger shares of Heavy edit may indicate lower translation quality, recurring reviewer corrections, or language-specific challenges. New text and Retranslation indicate that reviewers created new translations instead of editing existing ones.
Categories include:
Accepted — translations accepted without edits.
Light edit — translations with minor edits, below 15% edit distance.
Medium edit — translations with moderate edits, between 15–30% edit distance.
Heavy edit — translations with significant edits, above 30% edit distance.
New text — translations treated as newly created text during review.
Retranslation — translations that were retranslated during review.
Larger shares of Heavy edit, New text, or Retranslation may indicate lower translation quality, recurring reviewer corrections, language-specific challenges, or review steps that include translation work.
Translator quality leaderboard
The Translator quality leaderboard compares translators by post-edit rate, average edit distance, and review outcomes.
Use this table to understand how translations created by each translator performed during review. Higher post-edit rate or avg. edit distance values may indicate translations that required more reviewer corrections.
The metrics are based only on translations that were reviewed, not on all translations created by the translator. New text and Retranslation indicate cases where reviewers created new translations instead of editing existing ones.
Columns include:
Translator — the contributor who created the translations.
Translations reviewed — the number of the translator’s translations that were reviewed.
Words reviewed — the number of source words reviewed for that translator’s translations.
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Avg. edit distance — the average edit extent as a percentage of segment length, calculated only from reviewed translations. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
No edits — the number of translations accepted without changes.
Translations edited — the number of reviewed translations that were edited.
Light edits — the number of translations with minor edits, below 15% edit distance.
Medium edits — the number of translations with moderate edits, between 15–30% edit distance.
Heavy edits — the number of translations with significant edits, above 30% edit distance.
New text — the number of translations treated as newly created text during review.
Retranslation — the number of translations that were retranslated during review.
Higher post-edit rate, avg. edit distance, or heavy edit values may indicate translations that required more reviewer corrections. Higher New text or Retranslation values may indicate cases where reviewers had to create or retranslate content instead of only reviewing it.
Results based on low reviewed translation volumes should be interpreted cautiously.
Reviewer performance
The Reviewer performance table compares review activity, translation effort, and turnaround time across reviewers.
Use this table to understand how reviewers handle review work, identify differences in review patterns, and spot potential workflow inconsistencies. Large differences between reviewers working on the same language or workflow may indicate inconsistent review standards, content complexity differences, or workflow variations.
The metrics are based on translations that were reviewed, not on all translations in the project. New text and Retranslation indicate cases where reviewers created new translations instead of editing existing ones.
Columns include:
Reviewer — the contributor who reviewed the translations.
Tasks — the number of review tasks handled by the reviewer.
Translations reviewed — the number of translations reviewed.
Words reviewed — the number of source words reviewed.
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Avg. edit distance — the average edit extent as a percentage of segment length, calculated only from reviewed translations. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
No edits — the number of translations accepted without changes.
Translations edited — the number of reviewed translations that were edited.
Light edits — the number of translations with minor edits, below 15% edit distance.
Medium edits — the number of translations with moderate edits, between 15–30% edit distance.
Heavy edits — the number of translations with significant edits, above 30% edit distance.
New text — the number of translations treated as newly created text during review.
Retranslation — the number of translations that were retranslated during review.
Avg. turnaround hours — the average time between translation completion and review completion.
Avg. review active hours — the average active review time spent by the reviewer.
Higher post-edit rate, avg. edit distance, or heavy edit values may indicate that the reviewer is handling content that requires more corrections, or that review standards vary across reviewers or workflows. Higher New text or Retranslation values may indicate cases where reviewers had to create or retranslate content instead of only reviewing it.
Translator x reviewer matrix
The Translator x reviewer matrix shows quality metrics across translator and reviewer pairings by language.
Use this table to identify collaboration patterns that may require closer review. Higher post-edit rate or avg. edit distance values may indicate collaboration mismatches, language-specific challenges, or quality issues.
The metrics are based only on translations that were reviewed, not on all translations created by the translator. New text and Retranslation indicate cases where reviewers created new translations instead of editing existing ones.
Columns include:
Translator — the contributor who created the translations.
Reviewer — the contributor who reviewed the translations.
Base language — the source language for the reviewed translations.
Target language — the target language for the reviewed translations.
Tasks — the number of tasks involving this translator-reviewer pairing.
Translations reviewed — the number of reviewed translations for the pairing.
Words reviewed — the number of source words reviewed for the pairing.
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Avg. edit distance — the average edit extent as a percentage of segment length. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
No edits — the number of translations accepted without changes.
Translations edited — the number of reviewed translations that were edited.
Light edits — the number of translations with minor edits, below 15% edit distance.
Medium edits — the number of translations with moderate edits, between 15–30% edit distance.
Heavy edits — the number of translations with significant edits, above 30% edit distance.
New text — the number of translations treated as newly created text during review.
Retranslation — the number of translations that were retranslated during review.
Higher post-edit rate, avg. edit distance, or heavy edit values for specific pairings may indicate review style differences, collaboration mismatches, language-specific challenges, or quality issues. Higher New text or Retranslation values may indicate cases where reviewers had to create or retranslate content instead of only reviewing it.
Task quality by project
The Task quality by project tab helps you analyze review quality at the project and task level.
Use this section to compare post-edit rates across tasks, understand how review effort is distributed by project, and identify projects or review tasks that may require additional context, quality improvements, or workflow adjustments.
Task post-edit rates
The Task post-edit rates chart shows tasks grouped by post-edit rate bucket.
Use this chart to understand how much reviewer editing was required across tasks in the selected period. Lower post-edit rates usually indicate that translations required minimal reviewer changes, while higher post-edit rates may point to quality, context, or workflow issues.
Post-edit rate calculations exclude cases where the reviewer created new text or retranslated existing content, since those represent translation work rather than review edits.
Buckets include:
<10% — very low post-editing effort; most translations were accepted with little or no editing.
10%–30% — low to moderate post-editing effort.
30%–50% — moderate post-editing effort.
50%–70% — high post-editing effort.
70%–90% — very high post-editing effort.
>90% — almost all reviewed translations required editing.
Use this chart to quickly assess the overall quality distribution of tasks and identify whether a significant share of tasks required heavy reviewer intervention.
Project review summary
The Project review summary chart shows review effort by project, grouped by edit intensity.
Use this chart to compare review patterns across projects and identify where additional quality checks, context improvements, or workflow changes may be needed. Higher shares of Heavy edit, New text, or Retranslation may indicate quality or workflow issues.
New text and Retranslation indicate cases where reviewers created new translations instead of editing existing ones.
Categories include:
Accepted — translations accepted without edits.
Light edit — translations with minor edits, below 15% edit distance.
Medium edit — translations with moderate edits, between 15–30% edit distance.
Heavy edit — translations with significant edits, above 30% edit distance.
New text — translations treated as newly created text during review.
Retranslation — translations that were retranslated during review.
Higher shares of Heavy edit, New text, or Retranslation may indicate quality, context, or workflow issues that require additional reviewer effort.
Detailed project review
The Detailed project review table provides project-level review quality metrics, including post-edit rate, average edit distance, edit distribution, and recent review activity.
Use this table to identify projects with higher review effort, quality issues, missing context, or workflow patterns that may require improvement. New text and Retranslation indicate cases where reviewers created new translations instead of editing existing ones.
Columns include:
Project — the project name.
Tasks reviewed — the number of tasks reviewed for the project.
Translations reviewed — the number of translations reviewed.
Words reviewed — the number of source words reviewed.
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Avg. edit distance — the average edit extent as a percentage of segment length. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Translations edited — the number of reviewed translations that were edited.
Light edits — the number of translations with minor edits, below 15% edit distance.
Medium edits — the number of translations with moderate edits, between 15–30% edit distance.
Heavy edits — the number of translations with significant edits, above 30% edit distance.
New text — the number of translations treated as newly created text during review.
Retranslation — the number of translations that were retranslated during review.
Last review — the date of the most recent review activity for the project.
Post-edit rate is color-coded to make project quality easier to scan: green indicates healthier values below 30%, while red highlights values above 50% that may need closer attention.
Review task detail
The Review task detail table provides detailed review quality metrics for individual review tasks, including post-edit rate, average edit distance, edit distribution, turnaround time, and review activity.
Use this table to identify review tasks with potential quality, context, or workflow issues that may require closer analysis. New text and Retranslation indicate cases where reviewers created new translations instead of editing existing ones.
Columns include:
Task — the task name.
Task ID — unique identifier of the task.
Project — the project associated with the task.
Language pair — the source and target language pair reviewed in the task.
Reviewer — the contributor who reviewed the translations.
Translations reviewed — the number of translations reviewed.
Words reviewed — the number of source words reviewed.
Post-edit rate — the percentage of reviewed translations that required edits. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
Avg. edit distance — the average edit extent as a percentage of segment length. Translations where the reviewer created new text or retranslated existing content are excluded from this calculation.
No edits — the number of translations accepted without changes.
Translations edited — the number of reviewed translations that were edited.
Light edits — the number of translations with minor edits, below 15% edit distance.
Medium edits — the number of translations with moderate edits, between 15–30% edit distance.
Heavy edits — the number of translations with significant edits, above 30% edit distance.
New text — the number of translations treated as newly created text during review.
Retranslation — the number of translations that were retranslated during review.
Reviewed turnaround hours — the time between translation completion and review completion.
Review date — the date when the review was completed.
Post-edit rate is color-coded to make tasks easier to scan: green indicates healthier values below 30%, while red highlights values above 50% that may need closer attention.
AI Scoring
The AI Scoring dashboard helps you compare AI-generated quality scores with actual review outcomes.
You can analyze AI Score alongside post-edit rate, edit distance, and reviewer behavior to understand how well AI scoring reflects the amount of editing required during review. This can also help you identify projects, languages, or workflows where your AI translation setup may need adjustment.
AI Score is a quality signal, not an absolute measure of translation quality. A high score does not guarantee that a translation will require little or no editing, while a lower score does not necessarily mean that substantial changes will be needed.
Filtering data
You can filter the data by date, project, target language, task, workflow, translator, and reviewer.
Combine filters to investigate whether a particular pattern is associated with a specific project, language pair, workflow, or contributor.
AI translations
This section shows how much AI-translated content is included in the selected data:
AI translations — number of AI-translated translations reviewed during the selected period.
Share of AI translations — percentage of all reviewed translations that were translated using AI.
AI post-edit rate — percentage of reviewed AI translations that were edited by reviewers. A higher value means that a larger share of AI translations required changes.
AI Score coverage — percentage of AI translations for which an AI quality score is available.
These metrics provide context for the rest of the dashboard. If the selected filters include only a small number of translations, interpret the results with caution.
AI confidence vs. review effort
This report compares AI Score ranges with the amount of editing performed during review.
Use it to see whether higher AI scores are associated with less reviewer effort. For example, if translations with higher scores are accepted more often and require fewer edits, AI scoring is generally aligned with reviewer outcomes.
If translations with high AI Scores still require substantial editing, use the available filters to investigate whether the pattern is associated with a specific project, language, workflow, or other part of your translation setup.
A recurring pattern of high AI Scores and high editing effort may indicate that your AI translations need further adjustment, for example to better account for terminology, style, tone, or other content requirements.
Avoid drawing conclusions from a small number of translations. Look for recurring patterns across a meaningful volume of reviewed content.
AI flags vs. review effort
This report compares AI-detected issue levels with the amount of editing performed during review.
Use it to see whether translations with more severe AI flags also require more reviewer effort. If higher issue levels are associated with higher post-edit rates or greater edit distance, the detected issues generally correspond to reviewer outcomes.
If reviewers frequently accept translations with AI flags, or substantially edit translations with few or no flags, investigate the pattern further. This may indicate a difference between the issues detected by AI and the changes reviewers consider necessary.
Issue severity and type
The issue reports help you understand what AI is flagging and how frequently different issues occur.
You can analyze:
Issue severity breakdown — how AI-scored translations are distributed across issue severity levels, including translations where no issues were flagged.
Issue type breakdown — the quality issue types most frequently identified by AI.
Issue types by severity — how identified issue types are distributed across severity levels.
AI flags not acted on — the percentage of AI-flagged issues that were not addressed during review, broken down by issue type and severity.
To learn more, please refer to:
A high rate of flags not acted on does not necessarily mean that the AI flag is incorrect. Use this metric to investigate whether AI scoring and your reviewers may be applying different quality expectations.
Issue severity over time
Use Issue severity over time to monitor how the distribution of AI-detected issues changes over time.
For each severity level, the dashboard compares the latest three monthly periods and summarizes the trend:
🚩 Worsening — the number of issues increased in each of the last two months.
⚠️ Caution — the number of issues increased in the latest month, while the previous month was stable or improving.
✅ Improving — the number of issues decreased in each of the last two months.
➖ Stable — no consistent upward or downward trend was detected.
Use these indicators to identify changes that may require further investigation. Interpret them together with translation volume, as an increase in the number of issues may simply reflect a larger amount of AI-translated content.
Top issues by language pair
The Top issues by language pair report shows the most common AI-detected issues for each language pair.
Use it to identify:
the most common AI-flagged issue types for each language pair;
which keys are affected;
issue severity;
average editing effort; and
how issue severity is trending over time.
Use this report to determine whether recurring issues affect AI translations more broadly or are concentrated in a particular language pair or subset of content.
How to interpret AI Score and review effort
The most useful question isn't simply “Is my AI Score high?” but “Does the AI Score match what my reviewers actually experience?”
AI Score | Review effort | What it can indicate |
High | Low | 🟢 Expected outcome. AI predicts good quality, and reviewers generally make few or no changes. |
Low | High | 🟡 Expected signal. AI identifies quality issues, and reviewers also make substantial edits. |
High | High | 🔴 Needs investigation. AI predicts good quality, but reviewers still make substantial edits. Look for recurring patterns by language, project, workflow, or issue type. |
Low | Low | 🟡 Potential calibration gap. AI flags quality issues, but reviewers generally accept the translations with little or no editing. |
Use AI Score together with post-edit rate, edit distance, issue flags, and reviewer behavior rather than treating any individual metric as a definitive measure of translation quality.
Note on post-edit metrics and review activity
Translation quality reports include two types of insights:
Post-edit metrics, such as Post-edit rate and Avg. edit distance, focus only on reviewer edits to existing translations. Translations where the reviewer created new text or retranslated existing content are excluded from these calculations.
Review activity metrics include all reviewed work, including cases where the review step involved New text or Retranslation. These categories help identify tasks where reviewers performed translation work instead of only reviewing or editing existing translations.
Use Post-edit rate and Avg. edit distance to evaluate pure post-editing effort. Use charts and tables that include New text and Retranslation to understand the full scope of reviewer activity and identify workflows where review may be turning into translation work.