When the MQM tab appears

The MQM tab appears in the job details view when the job is authorized (In Progress, Completed, or Closed) and uses a workflow that includes at least one workflow step that:

1. Is LQA-enabled (meaning Use this step for Linguistic Quality Assurance is enabled in the workflow step configuration, and
2. Uses an MQM-compatible schema

If a job meets these conditions, the MQM tab appears alongside the other job details tabs. If the job does not use a workflow with an MQM-enabled LQA step, the tab does not appear.

The data in the MQM tab is not limited to evaluations completed by LQA Agent. It appears for any job where MQM-based quality evaluation occurs, whether LQA errors are recorded by LQA Agent or by a human evaluator.

What you'll see

The MQM tab provides a detailed breakdown of MQM results based on LQA reviews completed by the LQA Agent or human linguists. The scores reflect translation quality as measured by your LQA schema.

At the top of the tab, you'll see the overall MQM Quality Score. Below that, the MQM Results table provides a breakdown by locale, showing quality ratings and the number of critical errors for each. Buttons also let you download results or quickly view strings with errors or strings marked as no error.

If any locale in the job has fewer than 1,000 reviewed words, a conditional information banner appears at the top of the tab because the sample size is too small to be considered conclusive. For more information, see The 1,000-word minimum and Inconclusive results below.

The Overall MQM Quality Score is a single score summarizing quality across all locales in the job. When the job is still in progress, the note Updates as job progresses appears beneath the score because some strings may still need to pass through the LQA-enabled workflow step and be reviewed.

The MQM Results table provides a detailed breakdown grouped by LQA schema. Because a job can use more than one workflow, multiple schemas may appear in the same job.

MQM Results table

The results table is organized by schema. Each schema section includes the schema name and its passing threshold (for example, "Passing threshold: 90").

Only locales that have been evaluated appear in the results table. Locales that have not yet been reviewed are not listed. As evaluations progress, locales are added once their results become available.

The table includes the following columns:

Buttons

Two buttons appear next to the MQM Results heading:

• View Strings: A dropdown menu with two options:
  • View strings with errors: Opens the Strings View filtered to show only strings that have recorded LQA errors.
  • View strings marked no error: Opens the Strings View filtered to show only strings that have been explicitly marked as having no errors.
• Download: Downloads the MQM results as CSV files. One CSV file is generated for each schema used in the job. The file name follows the format: projectName_jobName_schemaName_MQM-results.csv. Each CSV includes the following columns: Project Name, Project UID, Job Name, Job UID, Schema Name, Schema UID, Passing Threshold, Locale ID, Word Count, Rating, Quality Score, and Critical Errors.

Multiple schemas

If a job's workflow references more than one LQA schema, the results table displays a separate section for each schema, each with its own passing threshold. This can happen when different workflows use different schemas for their LQA-enabled steps.

Understanding Reviewed Words

The Reviewed Words column in the MQM Results table shows the Reviewed Words count, which may differ from the job's total word count.

This is because the Reviewed Words count reflects only the strings that have undergone LQA review. Specifically, it counts strings that either:

• Have had LQA errors recorded against them, or
• Have been explicitly marked as no error.

Strings that have not been reviewed through LQA are not included in the count.

In practice, this means:

• If only a portion of the job's strings pass through an LQA-enabled workflow step (for example, when a job is split between two workflows and only one includes LQA), the Reviewed Words count will be lower than the total job word count.
• For LQA Agent evaluations, the Reviewed Words count typically populates quickly because LQA Agent processes strings automatically. For human evaluations, the count increases incrementally as reviewers progress through the job.

The 1,000-word minimum and Inconclusive results

MQM scores are calculated using a Reference Word Count (RWC) of 1,000 words. This standard MQM practice normalizes scores across projects of different sizes so they can be compared consistently.

Because of this, quality evaluations based on fewer than 1,000 reviewed words are considered less statistically reliable. To account for this:

• Locales with fewer than 1,000 reviewed words are flagged in the MQM Results table with a warning triangle icon next to the Reviewed Words count.
• These locales receive an Inconclusive rating instead of Pass or Fail. The quality score is still shown so you can interpret the results, but Smartling does not assign a definitive rating because the sample size is too small to be conclusive.
• When any locale in the job has fewer than 1,000 reviewed words, a blue information banner appears at the top of the MQM tab stating: When fewer than 1,000 words are reviewed, locales are rated Inconclusive rather than Pass or Fail.

Tip: Smartling recommends a minimum of 1,000 words in any LQA sample. When configuring automated sampling, ensure your sampling settings produce samples that meet this threshold.

MQM results are tied to the job where they were recorded

MQM error records are permanently associated with the job where the quality evaluation took place. This is important to understand if your workflow involves moving strings between jobs.

When a quality error is recorded by LQA Agent or a human reviewer, the error is linked to the job that was active at the time of evaluation. If you later move strings into a different job, the MQM errors do not move with them. The original job retains the error records, and the new job's MQM tab will not display errors recorded in the original job.

In most workflows, this is not an issue because quality evaluation and job completion happen sequentially. However, if your workflow involves moving strings between jobs after LQA has already occurred, keep in mind:

• The original job's MQM tab continues to display the quality results, even if the strings are no longer in that job.
• The new job's MQM tab only displays results from evaluations completed within that job. It does not inherit results from previous jobs.

Review pending

If the job qualifies for the MQM tab but no LQA evaluations have occurred yet, the tab displays a Review pending message.

Related resources

• LQA: Overview
• LQA: Setup Guide
• LQA Errors & Arbitration Report
• LQA Suite: Automated Sampling