Multi-country research shouldn't wait on translation.
We build translation pipelines for research teams. Your export comes back translated and checked, in its original format. A linguist only reviews what gets flagged.
In market research, translation sits on the critical path. Before fieldwork, and after it.
Before fieldwork. The questionnaire is signed off in English, then waits for its translations before each market can be scripted, checked and launched.
After fieldwork. Open ends come back in every language at once, and coding can't start until someone can read them.
On qual projects. IDIs and groups have to be transcribed and translated before analysis begins.
Wave after wave. The same questions and the same scales are translated again, and not always the same way.
Most segments are routine. The few that aren't are where a study breaks.
How it works
One file in. The same file out, translated and checked.
A pipeline set up for your team, your tools and your terminology. It is not a self-service tool: we build it, we run it, we maintain it.
01
Your file
as exported
02
AI translation
your glossary, your past translations
03
Second pass
against the source
04
Automated checks
tags, numbers, scales
Passedstraight through
05Flaggedto a linguist
06
Your file
original format, with change log
01
Your file goes in as it is
Survey platform export, open-end file, interview audio or transcript. No reformatting on your side.
02
AI translation, properly briefed
It works from your glossary, your past translations and your instructions, not from a blank page.
03
A second pass against the source
A separate review reads every translation against the original, then corrects it or flags it.
04
Automated checks
Tags and piped variables intact, numbers unchanged, scale wording identical from one wave to the next, nothing left untranslated.
05
A linguist reviews what was flagged
Human attention goes where it is needed, instead of being spread thinly across every segment.
06
The file comes back in its original format
Ready to load back in, with a log of everything that was changed.
Four uses
From survey translation to reports: what goes in, what comes out, what a human still reviews.
01 · Before fieldwork
Questionnaires
In
The master questionnaire, as a platform export or a translation overlay.
Out
The same file for each language: tags and piping intact, scales worded as in previous waves, ready for scripting.
Human review
A native read before launch is still recommended. A badly worded question can't be fixed once fieldwork is live. The reviewer starts from a checked draft and a list of flagged segments.
02 · After fieldwork
Open ends
In
Verbatims in the languages of the study, straight from the data export, respondent IDs untouched.
Out
An English translation next to each original, and a first grouping by theme as a starting point for coding.
Human review
Close to none. Open ends are read for meaning, not published. A linguist looks only at the responses flagged as ambiguous.
03 · Qualitative
IDIs and groups
In
Interview recordings or existing transcripts. The discussion guide helps with names and terminology.
Out
A checked English transcript, with the uncertain passages marked.
Human review
A linguist listens to the uncertain passages only: crosstalk, poor audio, names, strong accents.
04 · Delivery
Reports and presentations
In
Decks, reports and toplines.
Out
The translated document in the same format, with terminology consistent with the questionnaire.
Human review
It depends on the reader. Flagged segments only for internal use; a full read by a native linguist for anything client-facing.
Who's behind it
Built by a translator, not by a software vendor.
Willy Paul is a professional translator. He has worked on survey questionnaires and open ends himself, and he trains other translators in post-editing machine translation. The pipeline checks for the errors that matter in research because he has had to catch them by hand.
He is also the co-founder of Timevo, an independent agency that builds AI agents and automations for companies, then stays on to maintain them.
Where machine translation breaks a study
A tag or a piped variable is altered.The script fails, or a respondent sees the wrong text.
A scale is worded differently from the last wave.The two waves are no longer comparable.
A number, a currency or a time frame shifts in passing.Nobody notices until the data looks odd.
A segment stays in English, or gets translated when it shouldn't.Brand names and answer codes are the usual suspects.
Trial
Try it on a study you have already run.
Send an old questionnaire or an old interview, ideally one you already have a translation for. We run it through the pipeline and send back the output, the flagged segments and the change log. You compare it with what you received at the time.
Get in touch
Tell us where translation slows you down.
Book a free 30-minute diagnostic: we look at one of your studies and at where translation holds it up, before fieldwork or after.