How to audit legacy translations for quality and consistency. Decision criteria for re-translation vs. TM-leveraged updates. Cost modeling for both approaches.
- Audit before deciding
- The decision criteria
- Sequence by exposure
Audit before deciding
The instinct on inheriting messy legacy content is to retranslate everything. That is usually the most expensive option and often unnecessary — much legacy content is fine, and you cannot tell which parts without looking.
Sample across sources and eras rather than reading top to bottom. Errors cluster by who produced the content and when, so a stratified sample finds the pattern quickly.
The decision criteria
Update when the content is broadly accurate and the problem is inconsistent terminology. A terminology pass across existing translations is dramatically cheaper than retranslation and fixes the visible symptom.
Rebuild when accuracy is unreliable, when the source has changed substantially, or when the content is compliance-critical and its provenance cannot be established. In regulated content, unknown provenance is itself a reason to rebuild.
Sequence by exposure
Do not rebuild the whole estate at once. Start with content that is customer-facing, compliance-relevant or high-traffic, and let the rest follow as it comes up for revision.
Every rebuild should feed the termbase, so later batches get cheaper. Rebuilding without capturing the decisions repeats the original mistake.