Target update process for tech docs
## Target update tracking process:
Target updates are any changes made by a reviewer or any other contributor to a translated file in any MR that is not a result of the source text being updated. In other words, target updates are changes (corrections) made to translations that are not done in Phrase. Target updates need to be reflected in the TM before delivering the same or similar (\< 101% TM matching) content in a future translation request.
Scope: Japanese tech docs only (for pilot). The agreed cadence for the pilot project is Weekly - Monday triage, Friday completion
1. The person making a change to the translation in an MR comments the change in the MR. (See example [here](https://gitlab.com/gitlab-com/localization/tech-docs-forked-projects/prod/gitlab/-/merge_requests/774#note_2910374654)).
2. The person then applies _The L10n-target-update:: New_ in the MR. This will allow Argos PMs to see MRs with target updates for adding to TM or B, QA (Phrase or QA scripting) or in the AI prompting.
3. Argos PMs filter MRs based on the _L10n-target-update:: New_ label in GitLab. See example of filtering [here](https://gitlab.com/gitlab-com/localization/tech-docs-forked-projects/prod/gitlab/-/merge_requests/?sort=created_date&state=merged&label_name%5B%5D=L10n-target-update%3A%3A%20New&first_page_size=100).{width="687" height="254"}
4. The Argos PM then goes through each MR with the _The L10n-target-update:: New_ label, by opening the MR and checking the Commits tab for commits made by a human team member and then reading the comments in the MR related to the change.
{width="696" height="126"}
5. If there is any uncertainty about the target update change in the comments, the Argos PMs tag a GitLab TPM (@mpehkonen) in a comment.
6. If any changes are symptomatic of a larger scale problem that benefits from more evidence, anyone from GitLab or Argos can add the ~"docs-post-launch" label to the MR and the ~"L10n-docs:: found-in-AI-MR" for AI specific problems. More labels will be added to as needed.
7. Once the target update contents are clear, the Argos PM then assigns the items to be implemented by Argos internally (in TM, TB, Engineering) and applies the label _L10n-target-update:: Triaged._
8. Items that cannot be solved by Argos (are not applicable to Argos processes and tooling), can be brought to the attention of GitLab engineers for engineering triaging by commenting on the issue and tagging the GitLab TPM.
9. Once the update has been completed the Argos PM or resolver applies the _L10n-target-update:: Applied_
_Any target updates done on an MR where the reviewer updated the TM (such as when a reviewer identifies a bug and fixes it in Phrase as well) should still be labeled with the L10n-target-update:: Applied. This is done to keep track of the number of target updates._
The purpose of the target update process is continuous improvement of translation quality and making sure that work done on translations outside the translation workflow in Phrase is preserved.
The labels will also give GitLab and Argos visibility into target-update-related:
* Workload
* Velocity
* Continuous improvement trend
This will be an item in the weekly Argos - GitLab meeting.
<table>
<tr>
<td>
**Activity**
</td>
<td>
**DRI**
</td>
<td>
**Consulted**
</td>
<td>
**Informed**
</td>
</tr>
<tr>
<td>
**1. Label application**
</td>
<td></td>
<td></td>
<td></td>
</tr>
<tr>
<td>
Apply _L10n-target-update:: New_ to MRs
</td>
<td>GitLab MR Reviewer</td>
<td>GitLab TPM</td>
<td>Argos team, GitLab PMs</td>
</tr>
<tr>
<td>
Apply the optional ~"docs-post-launch" and ~"L10n-docs:: found-in-AI-MR" labels.
</td>
<td>GitLab MR Reviewer</td>
<td>GitLab TPM</td>
<td>Argos team, GitLab PMs</td>
</tr>
<tr>
<td>
**2. Filtering & Review**
</td>
<td></td>
<td></td>
<td></td>
</tr>
<tr>
<td>
Track MRs with _L10n-target-update:: New_ label
</td>
<td>Argos PMs</td>
<td>
\-
</td>
<td>GitLab PMs, Implementation teams</td>
</tr>
<tr>
<td>
**3. Triage & Prioritization**
</td>
<td></td>
<td></td>
<td></td>
</tr>
<tr>
<td>
Determine implementation approach (TM/TB/QA/Engineering)
Apply the _L10n-target-update:: Triaged_ label for updates that only need TM updates.
</td>
<td>Argos PMs</td>
<td>Argos Technical leads from each area, GitLab TPM when needed</td>
<td>GitLab team</td>
</tr>
<tr>
<td>
**4. Assignment & Tracking**
</td>
<td></td>
<td></td>
<td></td>
</tr>
<tr>
<td>Assign to appropriate resolver (TM/TB/Engineering/QA)</td>
<td>Argos PMs</td>
<td>Implementation team leads</td>
<td>GitLab team</td>
</tr>
<tr>
<td>
**5. Implementation**
</td>
<td></td>
<td></td>
<td></td>
</tr>
<tr>
<td>Implement fixes in TM/TB</td>
<td>Argos linguist</td>
<td>Argos PMs, QA if needed</td>
<td>GitLab team, other Argos teams</td>
</tr>
<tr>
<td>Implement QA updates (Phrase/scripting)</td>
<td>Argos engineer</td>
<td>Argos PMs</td>
<td>GitLab team, other Argos teams</td>
</tr>
<tr>
<td>Implement AI prompting updates</td>
<td>Argos AI engineer</td>
<td>Argos PMs</td>
<td>GitLab team, other Argos teams</td>
</tr>
<tr>
<td>
**6. GitLab Pipeline Items**
</td>
<td></td>
<td></td>
<td></td>
</tr>
<tr>
<td>Identify items requiring CI/CD changes (cannot be fixed by Argos)</td>
<td>Argos PMs</td>
<td>GitLab engineering team</td>
<td>GitLab team/PMs</td>
</tr>
<tr>
<td>Implement CI/CD pipeline updates</td>
<td>GitLab Engineer</td>
<td>GitLab team</td>
<td>All stakeholders</td>
</tr>
<tr>
<td>
**7. Verification & Closure**
</td>
<td></td>
<td></td>
<td></td>
</tr>
<tr>
<td>Verify implementation completeness</td>
<td>Assigned Resolver</td>
<td>Assigned resolver</td>
<td>GitLab team</td>
</tr>
<tr>
<td>Apply L10n-target-update:: Applied in TMS label</td>
<td>Assigned Resolver</td>
<td>Argos PMs (for validation)</td>
<td>GitLab team, all stakeholders</td>
</tr>
</table>
## **Automated target update tracking - future automation project via Argo (Summarized from** https://gitlab.com/gitlab-com/localization/docs-site-localization/-/issues/683)
### **Vision**
Build a systematic, automated workflow to capture translation improvements made in GitLab MRs and automatically feed them back into the translation workflow for improving translation assets (TM, TB, QA scripts, AI prompts), eliminating manual tracking and preventing quality regression.
### **Core problem**
Translation fixes made during GitLab review aren't systematically captured in upstream translation assets, causing:
* Repeated manual fixes across localization cycles
* TM/TB inconsistencies
* Lost institutional knowledge
* Wasted reviewer time on recurring issues
### **Critical Context from Yan**
"Target updates affect multiple downstream processes: AI/MT engines, Translation Memory, Terminology, Human Review workflows, Pipelines - recurring errors that could have been solved upstream"
This aligns with **"outsource first"** - minimize internal automation to critical pipeline blockers while establishing clear communication channels for all target updates between GitLab and Argos.
### **Automation Roadmap**
**Phase 1: Manual Prototype (Current - Q4 2025)**
**Status:** In progress with labels created by GitLab team members
* Three-label system: L10n-target-update::New → L10n-target-update::Triaged →
L10n-target-update::Applied
* Manual application by GitLab team → Manual application by Argos →
Argos resolution → Manual application by resolver
* Metrics tracking: volume, velocity, false positives, time-to-resolution
**Phase 2: Full Automation through Argo**
* Automatic label application on detected target updates
* Argo dashboard integration for backlog visualization
**Next Steps**
* Deliver demoable practical prototype with real examples (Q1/2026)
* Refine detection criteria based on prototype learnings
* Plan pipeline-based automation implementation
### Links
Issue for gathering samples of target updates.
* Define target updates use cases in tech docs https://gitlab.com/gitlab-com/localization/docs-site-localization/-/issues/55+
Link to meeting resources 2025-12-10:
* https://drive.google.com/drive/folders/1Igf9B0hjsdqpefmP69zlhYlGSCzL9mSd?usp=drive_link
Link to meeting resources 2025-11-06:
* https://drive.google.com/drive/folders/1eCujCMXKFDt386QIfiZLSZIeIkqkFobE?usp=drive_link
Feature request for Tech Docs Dashboard
* https://gitlab.com/gitlab-com/localization/tech-docs-dashboard/-/issues/5
Future-facing proposal for automated process:
* https://gitlab.com/gitlab-com/localization/docs-site-localization/-/issues/683+
issue