The Edit Distance Report · 27 million sentences
Where your team's review time matters most
Budgets are flat and content keeps coming. We measured 27 million sentences to see where review still needs your team.
Free, no form. 14 sections, PDF, 0.8 MB.
Share of the text people change before it ships
47% less editing. Context-sensitive suggestions draw on your translation memory, glossary and past corrections.
84/100
AI suggestions go out unchanged.
98/100
Sentences pass review unchanged.
42→16%
Editing, month one to month ten.
9%
Editing on English to US Spanish, the busiest pair. Italian gets 26%.
The person who runs localization
Your role matters more than ever
Editing a suggestion is only 7% faster per word. The time comes back in the reading, so the work moves to deciding what gets read, reviewed and kept. Whoever runs that flow decides how much time AI saves.
You run the flow
- 1SuggestAI suggests from your memory and glossary. 84 of 100 go out as written.
- 2ReadYou choose what gets a full read. Smells flags where meaning may have drifted.
- 3ReviewYou choose which passes run. Keep them where the risk is, like legal and campaigns.
- 4KeepYou decide what the memory and glossary keep. As they fill up, editing falls.
Editing falls as the memory fills up
How it was measured
- Sample
- 27,040,017 sentences on the wxrks platform, August 2024 to September 2026.
- Change
- Words replaced, added, deleted or moved before a person confirmed the sentence.
- Privacy
- Group numbers only. No client named, no text read.
A note for your VP
Ready to paste into an email or a planning doc.
Where our translation review time goes
wxrks measured 27 million translated sentences. People change about 8% of AI suggestions and 15% of machine translation, and review leaves 98 of every 100 sentences as they are. The time we win back depends on what we choose to read and review.
The report is here. resources.wxrks.com/papers/context-sensitivity-research
What the data can't tell you
- It counts changes, not quality. Quality is the next study.
- It's production data, not a lab test.
- The learning curve follows accounts that stayed.