The same correction for the fifth time
Your French reviser has corrected the same translator's use of a client's product name on every job this month. Each time, the reviser sighs, fixes it and delivers. The translator never finds out, because the corrected file goes to the client and the translator's version sits in an archive. Next month, the same correction again.
Meanwhile your vendor manager is asked which French translators are strongest for legal content and has no data to answer with, only impressions.
Why feedback gets lost
- Revision is done in the CAT tool or in tracked changes, and the translator is never sent the result.
- Revisers have no quick way to say why something was changed.
- There is no common set of error categories, so feedback is inconsistent when it happens.
- Quality information is not recorded against translators, so vendor decisions rely on memory.
- Corrections that should update the termbase or TM are not captured.
What it costs
Revisers spend time on repeated corrections that should have been solved once. Translators do not improve as fast as they could, and some are unaware there is a problem at all. Vendor choices are made on impressions. Quality issues that reach clients cannot be traced to a cause. And if your agency works to a standard such as ISO 17100, evidence of revision and feedback is thin.
| Aspect | Today | With revision capture |
|---|---|---|
| Corrections | Made silently | Recorded with category and severity |
| Translator sees them | Rarely | Summary after each job |
| Quality over time | Impressions | Scores per translator and client |
| Term corrections | Lost | Proposed for the termbase |
| Evidence for audits | Hard to produce | Recorded per job |
How we build the feedback loop
- Revisers work in the CAT tool as now. Changes between the translator's version and the revised version are captured per segment.
- Revisers tag significant changes with a category, such as terminology, accuracy, grammar, style or client preference, and a severity, using a short list your lead linguists agree. Minor preferential changes can be left untagged.
- After each job, the translator receives a readable summary: the segments changed, the categories and any reviser comments.
- Translators can reply or dispute a correction, which goes to the reviser or a lead linguist, so feedback is two-way.
- Quality scores per translator build up over time, by client and subject, based on the categories and severities your linguists define.
- Terminology corrections are proposed for the client's termbase, and final revised segments go into the master TM.
Scores are there to support conversations and vendor decisions, not to replace a lead linguist's judgement about a translator.
What changes
The categories matter more than the tool. We work with your lead linguists to agree a short list that revisers will actually use, and a clear line between an error and a preference. Without that agreement, scores measure revisers' tastes as much as translators' work, and translators lose trust in the feedback. With it, the record is something both sides accept.
Translators learn from corrections, so the same errors stop appearing. Revisers spend more time on real quality issues and less on repeats. Vendor managers have evidence when choosing translators for demanding clients. Terminology corrections become part of the termbase. And if a client or auditor asks how you manage quality, you can show a clear record per job.
Good translators tend to value this more than anyone. Most want to know what their revisers changed and why, and rarely get the chance.
Is feedback lost at your agency?
- Revisers correct the same errors repeatedly.
- Translators rarely see their revised files.
- Vendor choices are based on impressions, not records.
- Corrections to terms do not reach the termbase.
- You would struggle to show evidence of revision feedback.